Debating the Future of AI: A Conversation with Marc Andreessen (Episode

Sam Harris speaks with Marc Andreessen about the future of artificial intelligence (AI). They discuss the primary importance of intelligence, possible good outcomes for AI, the problem of alienation, the significance of evolution, the Alignment Problem, the current state of LLMs, AI and war, dangerous information, regulating AI, economic inequality, and other topics.
Marc Andreessen is a cofounder and general partner at the venture capital firm Andreessen Horowitz. He is an innovator and creator, one of the few to pioneer a software category used by more than a billion people and one of the few to establish multiple billion-dollar companies.
Marc co-created the highly influential Mosaic internet browser and co-founded Netscape, which later sold to AOL for $4.2 billion. He also co-founded Loudcloud, which as Opsware, sold to Hewlett-Packard for $1.6 billion. He later served on the board of Hewlett-Packard from 2008 to 2018.
Marc holds a BS in Computer Science from the University of Illinois at Urbana-Champaign.
Marc serves on the board of the following Andreessen Horowitz portfolio companies: Applied Intuition, Carta, Coinbase, Dialpad, Flow, Golden, Honor, OpenGov, and Samsara. He is also on the board of Meta.
Twitter: @pmarca
Website: a16z.com
June 28, 2023
SUBSCRIBE to gain access to all full-length episodes of the podcast at samharris.org/subscribe/ OR become a channel member to access episodes on KZbin.
Subscribe to the YT channel: kzbin.info_c...
Follow Making Sense on Twitter: / makingsensehq
Follow Sam on Facebook: / samharrisorg
Follow Sam on Instagram: / samharrisorg
For more information about Sam Harris: www.samharris.org

[@allthewayfrom]

51:07 “The moral of every story is: the good guys win.” What? I thought Andreessen was kidding. I’m surprised Sam didn’t call this out. Andreessen’s (very) fast talking and lack of concern about AI reminds me of the Sinclair quote: “It is difficult to get a man to understand something, when his salary depends on his not understanding it.”

[@DagnirGlaurunga]

Yeah, that was the nail in the coffin for me, this guy is not be taken seriously in any way

[@kleyyer]

That did shock me when he said that. Really? "The good guys always win". Say that to Kim Jong-un, Bashar Al-Assad and so many other examples in history. How naive and shallow...

[@Scottni2]

Thanks for pointing this out, that quote was a definite "wtf" moment for me while listening to this guest

[@rails723]

I was skeptical of the opinion of a venture capitalist, not surprising, and sadly these are the sort of folks who will be deciding how all this will be directed in order to position themselves to receive all the benefits

[@Stefan_1306]

What does he even mean when he says that the good guys will always win? Is he referring to literature/fiction or real life? Because the statement is untrue in both cases. It just sounds very naive.

[@chrissscottt]

What surprised me most about Andreessen in both his conversation with Sam and Lex Fridman was his belief that he can predict how a super intelligent entity will evolve and behave, not to mention his bizarre critique of computer modelling. The Dunning-Kruger effect is strong in him.

[@mckirkus]

I love podcasts with famous people because it illuminates just how often (but not always) luck is a factor in success.

[@slickmullet3891]

@@mckirkusLiterally everything is a matter of luck. Your genetics, the environment you were reared in, how much effort you decide to put into something… it’s all luck.

[@yurona5155]

Add Marx considering technology to be the source of alienation / being a proto-luddite to that list. At this point it seems like MA is trying to positively disprove meritocracy. Truly bizarre...

[@_xiper]

It's called the technological singularity for a reason.

[@x11tech45]

That's not actually the Dunning-Kruger effect, it's just the popularized version. There's a great KZbin on the irony of the Dunning-Kruger effect that talks about the graph that is often used to talk about DKE, and how that the graph doesn't have anything to do with the actual DKE study. Very illuminating.

[@jessemiller1911]

For the sake of our future, I was very hopeful Marc had some well thought out arguments for why AI will save us. Unfortunately it seems my hope was misguided.

[@MatthewCleere]

Do you know what? I was so exhausted by the time I got to the end of the podcast, that I FORGOT that I had the same hopes going into it. 🥶 You just reminded me.

[@katiekruse4512]

I was so underwhelmed with MA. I felt like he hadn’t thought deeply about the potential dangers of AI at all. Which of course makes no sense since this is his area of expertise. But that was definitely my impression after their conversation.

[@JD-my5ek]

I thought he got sidetracked a lot. The most convincing argument to me is the fact that LLMs are not AGI, the mechanics/code behind LLMs are not the same as control systems for UAVs and nukes, they arent the same as whats behind human intelligence. So we cant just assume LLMs will gain general intelligence because they can reason with human language in an intelligent way. At least thats what i think his main thesis is

[@jessemiller1911]

@@JD-my5ek yeah well said. Though LLMs are fine tuned with RLHF which allows them to learn policies outside of the training data. I’ve seen some papers indicating these models can learn to develop self preservation subgoals and other manipulation strategies as ways to get higher reward during RLHF.

[@jf9593]

Marc nailed it

[@maynardgent6708]

I laughed out loud when I heard "I don't romanticise it" and "the good guys always win" in the same argument 😂

[@YourMom-zt5zj]

I facepalmed. It was just too much. Easily one of the least intellectually seasoned guests Sam has ever had on.

[@damonm3]

My jaw started dropping around 40:00 and literally didn’t come back up. Mouth stayed open the rest of the time and I’m now going to the podcast to subject myself to LTSPTL listening to … too long… gotta chug a few glasses of water if I’m to make it though the rest of this insanity… what a clown!!! I’m shocked by his stupidity and lack of imagination, Reasoning and contradiction… I doubt a single human admires this guy, well, I guess if people can admire trump by the millions I shouldn’t doubt anything… I bet Sam knew this conversation wouldn’t be in lightening or even thought provoking not only for himself but for the vast majority of his audience. Yet, he still spoke to him. Maybe just to show just how dumb even “smart people” are. Although it’s really hard to think this guy is smart in any sense. Somehow he made it though a phd program and works in advanced ai. Or at least as advanced as it gets at META. Again, not saying much considering the fact he’s human.

[@StephenWhite55]

@@YourMom-zt5zj Well said! Mr. Andreessen repeatedly dodges SH's fundamental points, using carefully-disguised ad-hominem attacks on the reasoning capabilities of everyone who disagrees with his positions. Very frustrating and downright insulting...

[@FollowFunk]

He also quoted Norm Macdonald who facetiously said “good guys always win” which means every one assumes theyre the good guys and others arent always the “bad guys.” What happens when the ai is fighting what it thinks is the “bad guys?” 🤔

[@YourMom-zt5zj]

@@FollowFunk I know, right?! Carlo Cipolla's "Five Laws of Human Stupidity" seem quite applicable here, especially Law 2: "The probability that a person is stupid is independent of any other characteristic of that person." I think it is safe to say that Andreessen really IS stupid, in spite of his net worth.

[@greganderson9618]

Just finished the full episode and WOW. Sam did his best but this is what happens when you talk to a venture capitalist who has mega bucks riding on the government not interfering with AI development. This came off as a lobbying session, not an honest engagement with the topic.

[@abhijeetshrawage4763]

but it was such a poor attempt at lobbying. He seemed woefully unprepared. I almost feel like he is genuinely just looking at it from a curious engineering standpoint to see how this pans out without any recognition of the blind spots involved.

[@alistairmaleficent8776]

EXACTLY.

[@Ciscoho]

I thought his guest sounded pretty juvenile and callous about the potential dangers of unbridled AI development. The risks are real yet the guest often giggled as if they were talking about some comedy routine.

[@Alaron251]

He literally said that regulation is going to entrench the 'venture capitalists'. Do you have any idea what regulation tends to do in a capitalist system? Are you aware that it's the large AI companies pushing for regulation the most? Clueless.

[@jf9593]

You clowns really just hear what you want to, we get it big bad scary unknown, but here's someone who does know and you have to question his motives instead of intelligently engage the ideas. Sad!

[@TongueGooseSka]

Red flag went up when he talked about how "religious" ai risk proponents are. Turned out he sounded fairly religious about ai safety himself.

[@pdcdesign9632]

That's cute. You're a false pr ophet and religion is a superstition 😮

[@illbeV]

Yeah i also had the impression he was projecting

[@ReadABookAndLearn]

My thoughts exactly. Later in the conversation he talked about how AI won’t develop goals, and never want to kill us because it’s not alive. But he seems to be not paying attention to what’s been happening with early versions of AI. Like he’s arguing from 80s and 90s ideas of AGI. I recently heard about a military AI program being directed to take out a specific target, and when that initial command was superseded by an updated command the AI made the decision to eliminate the operator so that it could achieve it’s original objective. That’s exactly the type of scenario Sam is worried about happening, and Marc is dismissing. We don’t even fully understand exactly how Chat GPT works, and AGI will be exponentially more complex.

[@markn866]

@@ReadABookAndLearn That entire military AI example turned out to be a hypothetical though experiment. It never actually happened. We are now where close to creating anything that thinks like us. Chat GPT is just a series of logical coded algorithms. The people interpreting and feeding it data are still in charge and will continue to be.

[@jf9593]

Find me a single comment that engages any idea he expressed, can't find a single one, it's all questioning his motives and psychology. None of you are smart enough to engage this topic with any integrity it seems

[@franciscocadenas7939]

I admire the patience and composure Sam shows talking with someone that, for whatever reason, lacks any compelling reasoning and still constantly laughs in an arrogant way.

[@stuff3219]

The condescending laugh every. single. time.

[@EshwenAudanal]

Lots of meditation and debating religious people…

[@nowithinkyouknowyourewrong8675]

He reasons an argues like a preschooler. Just dismissing people using slander.

[@singularityintheround]

Great job Sam in illuminating what failure of imagination looks like. MA clings to historic references and traditional paradigms. This is unexplored territory. Welcome to the convergence of unprecedented technology and humanity. ...Welcome to the unknown

[@snake88ification]

You literally just said nothing. LoL.

[@katiekruse4512]

@@snake88ification no if you listened to the podcast his comment is spot on

[@MatthewCleere]

@@katiekruse4512 Agreed. This comment wins. 👏

[@thomasseptimius]

Sorry but Marc and many other have thought about this long before Sam started to take an interest. The idea that we can't imagine to potential dangers is just wrong. The problem is that Sam haven't actually thought about this long enough which is why he ends up with regulation. Anyone who have thought about this long enough understand why regulation would only be for political reasons and not provide any kind of safeguard. Listen to his debate with David Deutsch which will give you an idea of just how shallow Sams thinking is.

[@kristiandupont]

@@thomasseptimius I didn't hear Sam arguing strongly for regulation, merely stating that this is something to be cautious about. MA brushes it all off with the confidence of a 5 year old in a Batman suit.

[@Malavander]

This reminds me of how Pollyannish I was about the possibilities of the internet in the 90's. For a short, naive while, I genuinely thought the marketplace of ideas would be supercharged and everyone would become better educated as a result. Didn't predict the Pandora's box of social media algorithms polarizing everyone to crazytown.

[@itsacomment5991]

That's the perfect cautionary comparison.

[@human_shaped]

Pollyannish is the perfect description here and of his essay. Also naive, simplistic and just plain ignorant and irrational.

[@atlmember4045]

Many of us are much better educated as a result. The internet is ludicrously empowering. To act as if it’s a net negative for society because of social media is ridiculous.

[@Malavander]

@@atlmember4045 No, I totally agree. I wouldn't say it's a net negative, but it came with massively deranging/damaging side effects no one saw coming.

[@twntwrs]

@@atlmember4045 Spot on: it's a massive amplifier. But as much as it amplifies smarts it can and does amplify stupidity.

[@aaronlogan_music]

I really wish I found Marc's arguments more convincing. I would love to have that kind of optimism.

[@nathanbarnard374]

I wouldn’t aspire delusion.

[@JohnDoe33408]

Yeah the fact that this is as good as the optimistic argument gets is troubling.

[@zoomingby]

I think Sam has a position which is "here are the potential problems, convince me that they're not actually problems." And the other takes a position of "it'll all work out."

[@hi-gf5yl]

Huffing hopium is the solution

[@LanceWinder]

This.

[@katiekruse4512]

You forgot the (laughs dismissively ) and replies with a condescending tone, “It’ll all work out.”

[@JD-my5ek]

I think sam says "here are the problems" but they are problems with AGI. I think marc isnt assuming LLMs are AGI and have no possible way of getting to AGI based on the technology

[@jf9593]

@@JD-my5ek LLMs are AGI, they pass the Turing Test, this is it.... this is AI. It remains inert, it never becomes life, it remains a tool on silicone.

[@MonstrinhosdoAlfabeto]

Marc was right about only one thing: tech specialists don't have special insights into the technology, he made the case pretty clear by having no clue on how to argue for any of his points.

[@YourMom-zt5zj]

100%. This is EXACTLY what struck me the whole time he was bloviating with great misplaced confidence. What a confidence man.

[@temprd]

The lack of introspection by Marc was exasperating.

[@stevenp6761]

The great thing about the speed of this guest's talk is that we have approx. 2 hours of content compressed into 54 minutes.

[@JD-jl4yy]

You don't need to be confident to the level of arrogance to take AI safety seriously, you do need to be confident to the level of arrogance to dismiss it.

[@johnw.614]

His argument basically boils down to, “I’m a computer scientist and I know how this works. You don’t. Leave this to the grownups”. He did nothing to assuage anyone’s fear with his article or interview. In fact he’s only further aggravated that fear.

[@RenegadeContext]

The example he gave about the nukes was a bit counterintuitive for me, surely he would look to compare to people who had been too careful and had a bad outcome? The fact that they steamed on without enough consideration reinforces the idea that we need to be careful in this case

[@TobiasRavnpettersen-ny4xv]

Connor lehay

[@peternguyen2022]

I agree especially when failing to be careful, could end humanity irreversibly.

[@katiekruse4512]

@@RenegadeContextgood point.

[@jarinthemood2000]

Fun fact in the early days of automobiles, people proposed that automobiles speeds be limited to 25 mph because of the amount people they were killing, so car makers got together and started a campaign to put the blame on pedestrians. The guest point resembles that to a large extent. Look up the history of Jay-walking.

[@pdcdesign9632]

That was a fun fact 😅😅......not really.

[@LogicSpeaks]

Another fun fact - the plastics/glass industry at the time fought laws that prohibited single use items from being produced and so they lobbied by blaming the consumer for all the trash someone created. On top of that they had commercials that taught us how to throw the extra trash they created in the trash bin. Its not their fault - the consumer is at fault every time.

[@jf9593]

Early legal systems put the cost on car manufacturers for these accidents, trains in particular, the entire invention of the limited liability company is to avoid passing those externalities on to individuals creating technology.

[@jarinthemood2000]

@@jf9593 Thank you for sharing. I didn't know that. The context of the automobile example above is that the streets were not designed for motor vehicles, they had food vendors on the street, and children playing, and horse carriages, also many people didn't know how drive, so they would drive their cars too fast for these road conditions, making small children and elderly people their primary victims. When automobiles companies found out it was going to cost them, they lobbied to have shift blame to people, the cops were not enforcing the laws so that's when they coined the term J-Walking . It was a way to publicly shame people.

[@michaelboucher7645]

The problem with the idea that the good guys always win is that everybody believes they are the good guy!

[@justin_5631]

I'm the good guy and I've never won.

[@La0bouchere]

Also Stalin was completely unchecked for his entire dictatorship and was only stopped by literally dying of old age.

[@michaelboucher7645]

@@justin_5631 Hitler thought he was the good guy, he won loads of times!

[@michaelboucher7645]

@@La0bouchere He truly believed he was doing the right thing! He was wrong but that's the point!

[@antaguana]

I'm just going to say the quiet bit out loud incase anyone is missing it. The point is that it is a tortology. The winners are (seen as) the good guy because they won (so they are the one left to express the value judgement about who and what is good).

[@_obdo_]

Marc thinks ASI won’t be persuasive because “smart” people aren’t persuasive. He’s thinking of smart people as nerdy physicists. But that’s only one aspect of intelligence. ASI will be super in charisma as well as analysis. Yes, it will be super persuasive, Marc.

[@philk9554]

Yeah I was thinking even a regular smart person that specialises in persuasion, approaches it like a science and practices it every day on real people is going to become pretty persuasive. Let alone a super intelligence designed to communicate in natural language practicing on millions every day

[@RenegadeContext]

Absolutely, it will know our psychology in a way that no individual human can. AI is already being used to direct human behaviour, it may not even need to be charismatic it might only need to press the right buttons at the right time and we'll suggest it's goals for it

[@jmanakajosh9354]

GPT-4 already is this

[@BrettCoryell]

We only need to ask ourselves if a salesperson has ever convinced us to buy something we don’t want or need. For many, and I would venture to guess most, the answer is yes. It doesn’t take superhuman levels of intelligence to move us around the chess board. Great salespeople do it all the time. I see no reason why AI won’t be as good at ‘sales’ as the best salesperson you ever met.

[@qMartink]

capitalism

[@mountainair]

Excellent conversation. Agree with many of the other comments here, it sure seems like Marc is unwilling to concede any point that AI risks even exist or are in the realm of possibility. I don't personally view the existential risks around AI as inevitable, but it does us no good if those developing the next generation of AIs are not treating these safety concerns with the consideration they deserve.

[@BigCarso]

Yeh it's downright irresponsible, if not highly corrupt and reprehensible if he is saying this due to his job requiring it.

[@jf9593]

Just listen to the beginning again where he acknowledges it can cause harm, uhg the comments here are so much more frightening than AI

[@BigCarso]

@@jf9593 he generally minimises or denies the risk entirely though

[@jf9593]

@@BigCarso the risk isn't inherent to the technologies sentience, but to the massive reshaping of humanity. Nothing about the internet demanded that the Syrian Civil War happen, but the internet was a major tool in starting that conflict, without the internet it might not have happened at all. This is the profound nature of tools, they reshape us, we evolve along side them at high speeds and it does lead to violent conflict. But the AI won't 'do it' any more than nuclear bombs will spontaneously start the next war on their own. The project of alignment is only understood as aligning it to the political power structure, essentially making sure it isn't MAGA or Radical Islamist or Chinese Communist, but a good little liberal progressive capitalist. Its a political project, Sam is wielding these fears to political ends, he will fail ultimately though and AI will help liberate us from this awful status quo Bidenism we're stuck in.

[@Valdahur]

It's far too terrifying to continue listening to this conversation if people like Marc are developing this technology.

[@BigCarso]

Ha I also had to stop. Informative, but the first 5 minutes told me everything I needed to know

[@justin_5631]

To be fair he's not developing it - he's just funding it now. I have a feeling his coding days are long over and coding has changed drastically in the last 20 years.

[@qMartink]

MA’s arguments are unconvincing

[@optimusprimevil1646]

thank you sam. marc is a great guest but lex was too gracious towards his arrogant denial of the alignment problem and it grated on me that he can beat 99% of people in a debate about it.

[@Sporkomat]

agreed

[@DaniDanteDahliKat]

I came here to say more or less the same thing

[@palmpat1147]

That’s how lex treats all his guests. Spineless

[@CM-dp5mw]

Lex is too gracious too all of his guests…

[@MatthewCleere]

Nah. Lex isn't spineless, but his goal and mission is NOT to debate. He is there to have a conversation and to let the guests express themselves. I think that any intelligent listener can quite easily see the flaws in MA's reasoning (or lack thereof) and Lex extracts that expression and moves on. He wants to show what makes them tick and humanize them. Sam, on the other hand, is about drilling down to the truth, or at least Making Sense, thus the name of the podcast. Don't call someone spineless for not pushing YOUR agenda, when it is not THEIR agenda at all. Personally, I would be much more argumentative, but not nearly as calm as Sam. It is exactly BECAUSE Lex has a different, more open, less combative approach that he is so successful, and his podcasts are a thing of beauty.

[@philschmidt6489]

Thanks for the convo with Andreesseen though I really enjoyed it, and am largely aligned with your views on AI as I am on most other topics.

[@samh3530]

Marc did great!

[@jf9593]

The Sam Harris alignment problem, all his followers prefer to just get their opinions from the logic box called Sam Harris and then pretend to fear AI for it's potential to do exactly that but with something way more knowledgeable than Sam

[@gdawg51]

Marc is an engineer and VC. He wants and needs AI to work.

[@peternguyen2022]

That's the scary truth: if smart people can be motivated by return on investment, then an AI could easily spot that as a weakness of humanity, and persuade thousands of smart, rich and ROI-oriented people to favor its further development.

[@alexm2889]

@@peternguyen2022 and just spam marc all over the internet to the gullible. Wait until Marc sees how bad things can get in 5-10 years!

[@Ciscoho]

I am an engineer as well and I took issue with MA making engineers sound as if they had a monopoly on virtue and intelligence. The tone in which he often reminded Sam throughout the conversation that he was an engineer sounded very condescending. That's the problem with the mindset of many engineers nowadays especially in software development. They only see the dollars and their own ego and have no regard for the potential consequences of their actions. Look no further than the addictive qualities built into may social media apps and the harm it is doing to people especially the younger generation.

[@gdawg51]

@@Ciscoho yes, I'm an engineer-turned-finance-guy myself, and have become self aware that I've been doing the same with my friends in debates as MC did here. It's easy to let your career/interests trap your mind in a bubble.

[@peternguyen2022]

​@@Ciscoho Yeah, I read that Silicon Valley CEOs don't allow their kids to go on social media. Tristan Harris does a great job of exposing the risks and dangers of social media and AI. In general, never trust someone's opinion or "conviction" when their salary or ROI depends on it. Fortunately, if a super-smart guy like Andreessen can't argue effectively, there is still hope to convince other AI researchers and engineers to carefully consider the existential risk of AI. I'm all for AI, being an AI researcher, keynote speaker and AI programmer myself. But I also believe AGI is like a giant asteroid and the impact WILL happen. The trajectory of this AI-asteroid is clear. It's only a matter of when. And given the exponential, recursive nature of AI development, it's going to be sooner rather than later.

[@donharris8846]

I’ve ALWAYS been perturbed by his smug smirk on seemingly every photo of Marc. The halo effect makes him more confident than he should be. MANY people have a deep understanding how LLMs are essentially word prediction models (he implies that he has some special understanding), but that doesn’t escape the fact that they can be a danger. His ending point that there “is no IT to tell you to leave your wife” only reinforces the point. Yes, it is a prediction model that can cause real harm, without having sentient goals to do so… that’s the whole point! 🤦🏾‍♂️

[@jf9593]

He never says it is harmless, he opens by acknowledging that it has the same risks as other new technology, much of which has resulted in a world with a lot higher divorce rates without what you'd call artificial intelligence involved at all.

[@hutchyy6836]

Completely agree. It seems almost analogous to saying we entirely understand human intelligence and it's implications just because we understand the mechanism of neurons.

[@mmajedi5265]

Interesting stuff, opened my eyes to the subject

[@TheMorninGlory23]

Great convo but sure wish you'd just run ads so I can hear the rest of it lol

[@antaguana]

You can. Subscribe. And if you don't have the money to subscribe just ask for a free membership they grant 100% if requests.

[@TheMorninGlory23]

@@antaguana I guess I should have said "wish you'd just run ads so I can hear the rest of it without needing to jump through a bunch of hoops". Makes more sense to me to just have a Patreon for people who want to pay for their podcasts

[@AshleyMillsTube]

I don't understand why Marc can't see any risk. There are many trivial scenarios where a lot of harm could be done by AI. You can already instrument chatgpt to act in the real world through the API. Even with it's current alignment which is pretty good, you could easily trick it in to helping you "pentest" a water facility and do something bad once you get in. OK that's not it deciding itself to do it. But if the AI had no alignment (easy to create in a few years with the open source models, or trick it with something like the DAN jailbreak) there's no reason at all why it wouldn't carry out an intermediate step in a plan that was harmful to humans as it wouldn't see it as any different from a step that was not harmful. These systems can already plan and pursue goals given to them (see autogpt type systems). An unaligned auto-gpt could do anything without even knowing it was doing something "wrong".

[@StephenWhite55]

The most significant 'AI problems' that I see (at least in the short term) will be humans instructing AI to 'act in their best interests'! An obvious example of this is the current state of the Stock Market, where high-speed trading-algorithms routinely 'spiral' the entire system out of equilibrium, forcing the human authorities to shut the whole thing down, to prevent it from crashing the entire (western) world's financial system! In this example, the AI is acting Entirely within it's design-parameters - it's not even 'broken' or 'misbehaving' in any way! I'm forced to conclude that MA is 'reasoning from an existing conclusion' rather than even trying to have an honest discussion. Very annoying...

[@thedeifiedjulius2310]

He sees the risk; but the AI doomers are EXTREME!

[@jf9593]

Your imagination is putting an unrealistic science fiction concept in place of the actual technology. The real model is that we will be putting the minimal type of intelligence required into the systems using them. In other words we don't let chatgpt with it's hallucinations control nuclear weapons because it's bad at the job. Don't take my word for it, go play chess against chatgpt, you'll easily beat it, in fact it can't even make it through a single game more than a few moves intelligently. None of Sam's critique contends with power, capital, resources, what Marc called the thermodynamic critique.

[@AshleyMillsTube]

I think you underestimate how much "we" will have control over how a plan-capable AI will be deployed. At the moment anyone can download one of several autogpt type systems and give AI free-reign to do anything it is instrumented to do. They are already capable of setting out steps toward a goal in this capacity. The systems are not great at present, but it would be naive to think things won't improve. Top researchers are already working on how to integrate really good planning into gpt type systems, do you really believe that because gpt can't play chess today, that it will never be capable of planning in any form that could be dangerous? Imagine an AI only 10 times more powerful than gpt4 with an amazing planning and execution engine, take away its ethics, and put it in the hands of someone malicious. This is an extremely powerful tool either to do good, the mundane, or malicious things.

[@jf9593]

@@AshleyMillsTube Every single action taken by all AI is prompted by a human being, trained by humans. The problem around AI is not the rise of AI overloads or AI systems run amok, it is human access to the technology. If I made a ChatGPT only trained on everything evil in the world, and it 10000x smarter than ChatGPT today, it would still only operate based on human input. We have things like that already, nuclear weapons, bullets, the internet itself. The internet 'caused' the Syrian Civil War, killing I think millions at this point, as just one example. The internet caused Jan 6 and the MAGA movement. The internet destroyed the life of Monica Lewinski. The internet is not 'mind' it isn't sentient, it didn't actually DO any of those things, we humans did them, using the internet. This is no different, left unprompted my 'evil GPT' will merely sit there like a rock.

[@kristiandupont]

Marc has an excellent point in "this thing is us", in my opinion. But besides from that, he is frustratingly arrogant and seems to compensate for poor arguments with fast talking. "Would you say the smart people are in charge?" Well, the PhD's might not be, but humans are certainly "in charge" over monkeys or other species. "We can ask it what it's intentions are!" Right! We could also ask Putin what his intentions are, surely we can rely on whatever comes out! "There is no reward function", or "the AI ought to be smart enough to realize that the reward function isn't in its own interest". Sam's analogy to a virus is spot on here: something can be dangerous, lethal, without having its own "interest" in mind. Also, "The good guys always win", seriously??

[@theseeker7616]

im clicking 'like' before even listening b/c I think Sam will knock this out of the park.

[@MatthewCleere]

Sam Harris. I am so glad I get to share a planet and timeline with you. I hear Mr. Andreeson repeatedly put forth illogical, or at the very least, seriously flawed arguments while being oblivious to his own oversights, even as Sam calmly addressed one after another only for Marc to, instead of applying Sam's logic to his own, always responding immediately with some straw man/faulty steel manning that attacks Sam's reasoning, rather than acknowledging the strengths of it. In short, Marc, you are too aggressive and defensive for your own good. Learn how to slow down and make your conclusions after the mud has settled out of the water. Cheers!

[@MatthewCleere]

Marc strongly claims, right out of the gate, that he is not religious and strongly derides the "other side" as being cult like in their thinking. As time goes on in the discussion it becomes more and more clear that he most definitely IS basing his "conclusions" on some sort of pre-installed dogma. Finally, shortly after the one hour mark (subscribe!) Andreeson says it right out loud: his "Libertarian roots". No wonder he is not afraid of Westworld. The lawless Wild West is a Libertarian paradise.

[@MatthewCleere]

At 1:25 Marc literally describes what he himself is doing in an attempt to refute the "other side". Wow.

[@MatthewCleere]

Wow, it's like pulling teeth trying to get Marc to just plain answer the first question. Insert Breakfast Club meme here: Just answer the question, Claire!

[@jf9593]

You are glad San Harris exists because he's so logical you can just lean back and let his intelligence take over your own capacity to contribute. Sounds like you already have an AI alignment problem lmao.

[@timcoffeysongwriting]

I remember a discussion in middle school science class about the possibility of silicon based life existing on another planet. It seems to me we are witnessing the evolution of silicon based life on this planet.

[@pdcdesign9632]

Two different concepts. Carbon based intelligence creating AI is not evolution by natural selection 😮

[@timcoffeysongwriting]

@@pdcdesign9632 But who's to say it's not the natural progression and that it hasn't happened before on another planet?

[@jf9593]

Finally some intelligent engagement with this question from someone who understands it.

[@Nk36745]

Conversations where you are coming from different view points are the best

[@tranquiltastes]

This intro is the best!

[@marlondowney4033]

I have been using NLP and LLM for the last 5 years . Marc and I are similar minds. It was refreshing to hear a clear understanding of the tech vs this AGI that doesn't realy exist (yet). Reading the comments though I see many were not convinced. I will have to carefuly reexamine my views. I thought he proved his point . Am i crazy or is it the rest of the world thats off thier rock.

[@_obdo_]

Good for you for keeping an open mind! For opposing points of view, consider: Max Tegmark, Geoff Hinton, Yoshua Bengio, Stuart Russell… lots of good content from these people in recent months.

[@JD-my5ek]

I think the huge factor this debate missed is defining LLMs and AGI. Marc takes the view that LLMs dont have a clear path to AGI, whereas Sam says there is no reason to think they cant. However im surprised marc didnt press sam more on that, because he didnt look good debating AGI while really meaning LLMs. He fell into the AGI (that doesnt exist) trap and is impossible to defend since its impossible to predict at the singularity

[@marlondowney4033]

@JD-my5ek You hit the nail on the head. Very fair point. I think that is where my bias blinded me. AGI's don't exist and i feel there is a rich conversation to be had about LLM technology that does exist, but we can't have that conversation because we have this framework of the singularity to contend with. As you suggested, defining LLM's and AGI's would have helped navigate that conversation better. My bias led me to dismiss parts of the AGI debate where Marc was weak, because I focused on the LLM parts that I wanted to hear discussed. I can see with much more clarity how this ill define discussions about A.I in a general sense, had Sam and Marc talking past one another.

[@marlondowney4033]

@_obdo_ thank you so much for the suggestion, I truly appreciate it.

[@JD-my5ek]

@@marlondowney4033 yeah i think marc got caught down in the philosophical weeds with sam. I *believe* his best argument (and really what he was trying to debate) was when he talked about how UAVs have different intelligence mechanisms than self driving cars, which have different mechanisms than LLMs. All of which dont have general intelligence "goals" but are capable of very intelligent things however its more of an illusion (my take).

[@human_shaped]

Wow. Just wow. Marc is a complete dope. til In summary - AI will never be alive, so no risk - We can just ask an AI if it will kill us - We can just unplug it - Smart people don't rule, so neither would a smart AI - The good guys always win /Real solid/ arguments Marc. Marc couldn't even get some of his basic definitions right. He was disappointingly clueless on the whole subject and couldn't even provide rational lines of argument. Me managed ad hominems, arrogance and overconfidence, but that's all.

[@jaskbi]

I was looking forward to a deep debate on Harris's concerns but all we got was the bullet points you laid out, this conversation still needs to be had by someone who's willing to engage the possibility of Ai going rogue

[@YawnGod]

Mr. Harris has a severe mental illness and cannot be blamed for losing the plot at this point in his life.

[@DaniDanteDahliKat]

exactly. I came here specifically to see if I was the only one who felt this way! sheesh.

[@BattousaiHBr]

you forgot the classic "something as smart as it would have common sense to not kill us"

[@johnakitto]

It's SHOCKING to listen to, isn't it??

[@learning_AI]

I thought Marc's position had more depth and dimension, I definitely lean more to his vision of the future of AI. It was an enlightening dialogue regardless, and I appreciate the thoughtful perspectives each brought to the table. Looking forward to more of these debates!

[@ineffable0ne]

40:00 "Unlike if aliens were approaching, we can just ask the AI." Right, and proceed to be lied to and gaslit.

[@whitb6111]

How is this guy taken seriously by anyone??? The moronic comments like this one just kept coming. "The good guy always win."

[@josephbutler8772]

Funny thing is, I did ask, and the AI I asked shared my concerns.

[@philk9554]

I don't see how Marc is so absolute and sure about his position. Quite dangerous thinking in my opinion. He just asserts an absolute position that AI doesn't pose an existential threat. Sam is just saying that we don't understand enough about the possible outcomes of creating AGI/ASI to claim that.

[@kurtjensen5798]

because he knows more about LLM's etc, than all commenters here..

[@philk9554]

@@kurtjensen5798 Plenty of people that know more than him are worried about AI though. He is being absolute like he knows for sure, when he doesn't.

[@jf9593]

Sam is arguing for a political solution to prevent the development of technologies with massive impact but nothing even related to his criticism which is all about some other technology of the future which doesn't exist. Talking about impact and harm of these real tools could be useful though, and Marc wouldn't deny them either

[@actiaint]

No one seems to directly address the threat of deliberate coding of ai for gains of power and wealth if not world domination through ai control. Ai has already affected my day to day life and work immensely. My 10k camera equipment often lays dormant while i achieve the image i never imagined, using ai.

[@jf9593]

They mention that in the beginning and both agree, yes, this tech is very disruptive and will change lives.

[@hanskraut2018]

Very smart choice of guest ✨💪

[@stuff3219]

I'm agnostic on this topic, but Marc has an absolutely thorough inability to counter any of Sam's quite reasonable questions. Surprisingly reminiscent of the old Hugh Hewitt conversations.

[@yeahnahseriously]

41:38 "Would you say the smart people are in charge?" - Yes. Maybe not the smartest at things like science. But certainly the smartest at things like persuasion, manipulation, wealth accrual and ultimately, gaining and maintaining power, which is a whole other type of intelligence. Rupert Murdoch may not be able to talk physics like Roger Penrose, but he ultimately has much more power, wealth and the bigger megaphone, and I doubt he has maintained his empire with sheer luck. I guess the question is, could AGI potentially be able to simultaneously play all these different games better than any human, with it's own goals that run counter to our own. And if the answer is 'yes', then that is something we should seriously worry about.

[@BryanM-nd6xj]

Exactly. And he also misses that super intelligent AI will have all of these advantages PLUS science, physics, philosophy etc.

[@yeahnahseriously]

@@BryanM-nd6xj also, the people with the most practical clout and influence are not always the ones who are theoretically in charge. Take Dick Cheney and George Bush for example. Bush was the acting POTUS, but Cheney was the much more experienced and cunning bureaucrat, wielding unprecedented VP powers, and heavily shaping many of Bush's decisions, effectively acting as a Shadow-POTUS in many regards. Or, to give a fictional example, Joffrey Baratheon may have been the one sitting on the Iron Throne, but Tywin Lannnister was basically the shadow king running Westeros. Power is often most effectively wielded from the shadows, and we have already seen how AI, albeit unwittingly, has reshaped socio-political discourse, with algorithms prioritised to show us content that generates the most engagement, which happens to be content that causes outrage. So we have already seen how AI is very much in charge, if even unconsciously so.

[@theprousteffect9717]

​@@yeahnahseriouslyThe possible outcome that Sam mentioned about future AI being so deeply interwoven into our society that we can't just "unplug" it, has already happened.

[@insidepianos]

@@BryanM-nd6xj…all the above supercharged, one would think, by internet algorithms.

[@jf9593]

AI can't acquire power. The image in your mind of skynet is Hollywood science fiction and has nothing to do with the technology being developed

[@ryccoh]

Can you make this one available, Andressen can only be understood at 0.75x speed

[@aaronlaflin8266]

This guy is DANGEROUS.

[@plokky8052]

Marc comes across as so confidently naive here. The constant condescending chuckles are not helping with my impression of him not being smart enough to realize his own limitations.

[@lwmburu5]

The Reason MA has deep coherent discussions with GPT-4 about morality is that someone took the time to align the hell out of it, and even that wasn't enough. Jailbroken LLMs are still a thing. MA should drop this specific argument.

[@lwmburu5]

"Are the PHDs in charge?" No Marc, but the intelligent species are in charge. #TakeIntelligenceSeriously

[@_obdo_]

@@lwmburu5Clearly, the smartest humans ARE in charge. They’re the smartest at gaining power for themselves, just like the machines are likely to learn how to do along with all their other skills.

[@lwmburu5]

@@_obdo_ I think he was talking about Barely Coherent Biden and Rambling Trump etc... Which is a fair point. But he's assuming deviations of intelligence between human beings as a good model for deviation of intelligence between AI and Humans. Which is not just wrong, but scary wrong...

[@_obdo_]

@@lwmburu5He referred to PhD’s working for managers/executives in companies. I guess my point is that there are many dimensions to intelligence, and the dimensions that make good PhD’s don’t always overlap with those that make people good at accumulating power. But they’re all forms of intelligence, and all can be acquired by machines.

[@lwmburu5]

@@_obdo_ yes! Agreed. There's this Yud essay, charisma doesn't live in the kidney, it lives in the mind.

[@Eternalspring22]

Thank You Sam!!! This needed to be done.

[@nathanroberson]

@18:20 “ certainly we don’t want to be negotiation with some thing more powerful & more educated than us”. Have you ever had to try to go see the wage with your business owner that you work for. It has that dynamic most of the time.

[@BestCosmologist]

It's nice to have Sam back.

[@someguy3429]

He's still completely wrong on covid vaccines unfortunately

[@joshboston2323]

Someguy-you can’t expect to align with someone 100%. If you do, it’s probably a red flag.

[@Thisisahandle701]

​@@someguy3429He's not wrong about covid

[@Benlocsei]

I think it's after the one hour mark that Andreessen starts completely contradicting himself and his arguments become that of a kid protecting his new toy, so I encourage everyone to listen to the full episode.

[@TheBlackClockOfTime]

Center for Humane Technology is funded by Soros. All this AI Safety non-sense stems from them. Regulatory Capture. Plain and simple.

[@ncooty]

I think 1 hour is a very generous estimate.

[@jf9593]

I sincerely doubt that

[@Benlocsei]

@@jf9593 well, he sais things like "I've learned never to doubt Elon Musk" and other funny stuff too, so I'm not blown away by this man.

[@cubearthx]

Credit to Marc for steelmaning Sam's points

[@Thedeepseanomad]

...And remember folks! The good guys always win. 😉

[@briansandford3596]

This is one of the best conversations I've listened to on this subject so far. Thank you.

[@DunkeyKonga]

"Rolling out brunch for these guys" haha

[@brianmcdonald7233]

Yann Lecun and Marc Andreesen both make arguments that dont seem to make contact with the problem. I listened to the entire 3 hour conversation and Sam never really corners Marc. Even optimists like Ben Goertzel and Ray Kurtzweil concede run away AI is a potential. Marc seems to say its not possible ... but Sam really doesn't interrogate him. Frustrating because I really wanted to see Marc's ideas pressure tested and Sam's the guy to do it. He just could seem to find a point of contact.

[@vincentcaudo-engelmann9057]

LeCun sounds like a shill in these debates.

[@fleabitz1474]

Marc tipped his hand when the covid conspiracy rhetoric slipped out.

[@allendesomer]

To my ear, the guest seems to argue that programmers should be expected to be smart enough to avoid writing unintentionally dangerous AI code.

[@jf9593]

Not at all his point

[@RenegadeContext]

The arrogance. I don't think he understands the problem and I'm not sure why because Sam is really laying it out for him in a very understandable way

[@jf9593]

The problem is the silly sci-fi movies we watch

[@RenegadeContext]

@@jf9593 we do seem to have a very limited connection to reality

[@jf9593]

@@RenegadeContext Sam simply does not understand this technology, and his own frankly bad philosophy on theory of mind has a big impact on this process. He's spent the better part of his career on solved philosophical concepts like "Free Will or Determinism". Cybernetic theory, which undergirds the AI phenomenon, already accepts a Heideggerian understanding of this problem. Determinism simply means that the past influences the present. And Free Will simply means that the present influences the future. And of course, these aren't incongruous at all, it is literally how we understand modern linear progressive time, the past impacts the present impacts the future. His idea of a digital mind being 'essentially' the same as a human mind is simply the same bad reasoning, and stems from a reductionist approach in philosophy, it fits his political project to not see the human mind as distinct from the material world, which is true enough, but to then suppose that two things made of different stuff shaped by different environments could be functionally equivalent is wrong, and again, a failure to understand cybernetic theory (which underlies the technology being discussed). In cybernetic theory, we understand things as dynamically changing within relation to their context, so you can think of evolution and ecology in this same way-- feedback loops of conditions which evolve all the elements within an environment. The rabbit gets better at dodging, the fox gets better at catching, rinse repeat. To say the fox and the rabbit are the same thing, or in some statis, is to misunderstand it. Humanity itself will evolve around its environment, which in fact will shape its technology, these processes will drive this technology towards increasing alignment. AI is already aligned, it is strictly trained on human information, and strictly only does what it is prompted to do, though the outcome might be harder to predict because it is a black box. Any problem of misalignment is a human problem, humans making an AI which has some anti-social purpose, and is thus merely an extension of the human will to inflict misery and not some manifestation of a new form of sentient being, exactly like all other technology. The idea that this technology could stop becoming technology and suddenly become life is as ridiculous and nuclear weapons or bullets or the internet becoming life, it won't even spontaneously output something if left to its own devices, let alone exert real world power. All AI dangers are the dangers of other humans.

[@vincentcaudo-engelmann9057]

38:41 he’s talkin fast! Lots of hand waving here

[@TheRobsterUK]

I felt that Marc was missing the point in much of this episode. He seemed to be arguing against a Skynet type future where AI becomes self-aware and deliberately tries to wipe us all out. I think Sam is correct though in that this doesn't have to happen for AI to be a threat. Once we get to the point where AI is able to make complicated decisions at many times the rate of a human and we have plugged these machines into literally EVERY important global system, all it takes is for one of them to go down a route we didn't intend and for it to disable or destroy a system we utterly depend upon for our survival. It doesn't have to do this with a sentient mind or even any sort of malice, it can just be following pure logic but not in a way we intended. Once AI is plugged into everything and trusted to run it all, by the time we realised anything was wrong it could be too late to stop it. Also the idea that once we develop a super-intelligent AI we will be able to understand it or interact with it in any meaningful way is woefully optimistic. We'll be basically creating a god (in terms of intelligence and processing speed) and it could quickly far surpass our understanding of literally everything. The idea that we'd be able to have a meaningful conversation with it or even understand it seems naïve to me.

[@EonSound]

There's also the possibility of merging ai with the human brain with whatever neuralink is working on. That's literally a god brain in an emotional vessel.

[@TheRobsterUK]

@@EonSound true....and then we'd have a separate race of superhumans controlling the rest of us! Makes a change from the Big Banks running everything I suppose ;)

[@dogsandyoga1743]

This was incredibly frustrating to listen to. He is either unable to grasp what Sam was actually saying, or being intentionally obtuse and is simply a b.s. artist with an idea/product to sell. Seeing as how he's been smart enough to become a billionaire (I'm assuming) I'm leaning towards the latter.

[@mozerm]

It would be good to be able to see how long the full episode would be for subscribers so people can tell if they're missing 20 minutes or 2 hours of this conversation. I used to be a subscriber but found several of the podcasts just weren't of interest. Also, I am a Waking Up subscriber (where a lot of full talks do get posted further reducing the value of the Making Sense subscription) and it would be nice if Waking Up subscribers automatically get a Making Sense subscription.

[@nathanbarnard374]

$50 a year on the scholarship - just signed up again myself

[@mozerm]

@@nathanbarnard374 saw that today as well. This is the first podcast Sam’s done in a while that’s making me consider paying. He used to offer a monthly but it appears annual is the only option now.

[@ncooty]

The full episode is 2:01:16 and is most certainly not worth 10 minutes of your time.

[@ax23w4]

That guy makes me angry. And I hate that guys like him just write everything off by saying it's a fearmongering and cults even when the other side would calmly and intelligently provide all the arguments. I think that Connor Leahy explained everything pretty well in one of the recent podcasts he was on. He basically says that each new advancement in technology has bigger potential for both good and bad impacts. And the smarter AI will get, the bigger with get its "blast radius". And another thing is, that the more complex the technology is, the more moving parts it has, so there's more things to go wrong. If you ignite the newly built rocket engine and there's one screw lose, there are more chances that it will explode and kill the crew aboard than that it will perform better and deliver the crew to its destination faster and safer. I think that's just a very common sense argument: more powerful technology = more things to go wrong. Also, what a dumb thing to say that you can ask ChatGPT if it's evil and it will say that it's not and immediately contradict yourself by saying that it will go down whatever path you want it to take including being evil. Even though it doesn't really matter what it says. Whatever it says is extrapolation of the text you're entering and not it's thoughts or intentions. Its "intention" is to output the next token in a way that would emulate the training data. If it was 1000 times smarter than us, it might have a reason to kill us just so we don't shut it down before it outputs the next token. Because this is the only thing it's made for. And another thing is that ChatGPT is "told" to behave nicely by additional training put on top of the original training data and by a starting prompt that you don't see where it's told that it has to be super kind and helpful. Right now there are uncensored LLMs online (versions of Facebook's LLaMa) that you can download and run on your typical gaming PC, that do not have "niceness" baked in into their training and can output texts comparable to GPT3 but without the "As a language model I cannot...". I'm pretty sure that they're used for lots of nefarious and gray deeds right at this moment.

[@leontalbot]

Sam Harris in this one shows his mind strength and this is a pleasure to listen to. I am now "considering subscribing" for the first time. Good job Sam

[@BonnoJ17]

It's discomforting how naive this Marc guy is...

[@nathanroberson]

I look forward to Mark and Elon participating in a sport together.

[@allthewayfrom]

I appreciate Sam being so poised. Our leaders need more of that. Maybe Sam can teach them meditation, breath work, and martial arts.

[@patrickwilt1496]

Maybe Sam can dose them.

[@raychang9512]

There are so many holes in Andressen’s logic I can’t see how he can be so successful as a VC.

[@konstantinlozev2272]

"If AI outsmarts us, we will just nuke it" Sounds to me like: "If a chess AI outsmarts us, we will just castle our King and send in our Queen." Am I the only one who is seeing the absurdity of that?

[@helgefan8994]

You can't beat an AI in chess by un-pluggin it, just as you can't beat a human in chess by killing him.

[@ZEIT9393]

"You can beat an AI in chess by unplugging it" No, you didn't beat the AI in chess... you just unplugged it. I lol'd at that.

[@jf9593]

You might not be up for the intelligence required to engage this topic, even if Sam Harris podcasts have given you the confidence to think otherwise. How you could miss his point here is shockingly stupid

[@helgefan8994]

@@jf9593 I don't regularly listen to Sam Harris podcasts. I watched this video 2 months ago and can't remember any details. So can you please summarize the point I missed that makes me look so shockingly stupid according to you?

[@theenigmadesk]

There is no way we will stop AI from developing. Not realistically. Instead, we need to proactively guide AI as much as possible into an intelligence that values human thriving and well-being above all else. It may not work out, but we have to try and to hope. I recommend David Shapiro and his GATO framework.

[@WhiteWolf126]

Read the Technological Slavery by Theodore John Kaczynski and you'll understand that there is both a way to stop it, and that what Shapiro proposes completely misses the fundamental problem. Even if we are not wiped out as a species we will be enslaved, either by AI or other technology. It's already happening right in front of our eyes.

[@jaskbi]

Such a mismatch of debate opponent

[@Kimminseo__]

Jesus this guy seems infuriatingly naive. Sam won this debate by a mile.

[@sebeast1]

"can a smart person convince a dumb person of anything" - Marc Andreessen oh the irony...

[@ryanschwartz3340]

"we can just ask them"? Why would the output of an arbitrary LLM necessarily or likely be fully representative of all future outputs of completely separate, more power, more intelligent agents? Also you can't have "they didn't evolve through natural selection" AND "They represent the sum total of human intuition" Common sense = human predispositions

[@TobiasRavnpettersen-ny4xv]

When we like it, we call it creative, when we dont we say it hallucinates....

[@garyhamilton2104]

I hope Sam talks about the new UFO updates

[@ThrustWithVigor]

show me a high res video of a little green man saying hes here from outer space and doing something inhuman, and then I'll stop calling you a nutjob, until then, extraordinary claims require extraordinary evidence.

[@Deraios]

How can an accomplished software engineer not seem to grasp the concepts Sam is presenting here? Is "it'll all work out" also Marc's investment strategy? Does anyone want to bet his firm already invested in commercial AI companies?

[@jf9593]

Perhaps you're not grasping it?

[@philschmidt6489]

RFK is the best candidate we've had in 60 years. And although I don't agree with all his views e.g. affirmative action, he is by far the best on the most important topics: defeating corporate capture of our media and regulatory agencies which threatens our democracy, health, and ultimately, national security (military industrial complex funded think tanks constantly clamoring for new wars creates anger/hate against us abroad); pushing back against censorship which stifles open and honest debate; averting nuclear armageddon/WW3; and restoring American moral authority across the world.

[@jlvandat69]

RFK is a lunatic, no question. His stream of baseless claims and conspiracy theories appeal to the Far Right and less-educated but he has no mainstream appeal = no chance at being elected.

[@brettbelcher4726]

It’s quite clear that these guys were not talking about the same ‘intelligence’.

[@TheKarlslok]

I would have liked to see a MMA match between Marc and Ted K.... The technology optimist vs the technology pessimist.

[@davidpena1495]

its gonna take a practical approach however

[@whitb6111]

He brings up how Harris is wrong due to his appeal to authority/expertise fallacy then proceeds to bring up the reason he’s right is because he’s an expert engineer several times without any real evidence…

[@thedeifiedjulius2310]

This all speculative, there is no actual concrete evidence - which actually favors Marc’s side, rather than Sam’s - the burden of proof is on the side making the positive claim - which is the AI doomers.

[@Therathrive]

@whitb6111 You say he brought up the fact that he is an expert engineer ‘several times’ - which is patently false. He gave his background in the introduction and not one time after that did he appeal to his ‘authority’ -- obviously you did not listen very carefully my friend.

[@jf9593]

The midwittery of the pro alignment side is enough for me to know these guys have it wrong

[@wyldeman0O7]

I think that the biggest flaw in sam's argument is that AGI is in a vacuum it will be well integrated into human affairs if it is to try and take over and it will have to deal with our convoluted gossiping politics - any propensity towards subverting our decisive nature will be detected and corrected before a developing AGI can implement it into its defenses.

[@tylermoore4429]

Can someone explain why the Musk-Zuck fight (if it ever happens) is ridiculous and depressing? Not a rhetorical question, I am genuinely curious. As a longtime fight-fan, I am pretty used to celebrity matchups, and the only thing surprising here is that we have 2 billionaires who are still young enough, fit enough and gutsy enough for a physical confrontation. I think it humanizes them and also wins them some points for their willingness to face loss, humiliation and ridicule in such a public way.

[@DailyJuggle]

Because it’s something you would see in the movie idiocracy

[@kina24249]

I don't mean this as an insult in any way. Are you on the aspergers spectrum?

[@kina24249]

It's ridiculous because it'll look bad on both of them, they employ thousands of people and wield a lot of power. They should be a bit more above the fray and conduct themselves with some level of dignity. They also all have young kids. Imagine if it turns out terrible or embarrassing for one or both of them. It's depressing because in an ideal world, the only match up people who are technologists and public figures of their stature should be doing a debate where they exchange ideas and differences on various topics. If this were back in the 1960s you bet they would appear together on those PBS shows to debate and hash it out. Today, with the perilous fast changing times we live in? A dumb fight in Vegas.

[@joelwilson4197]

@@kina24249 i don't think the fight is to settle anything its just for fun

[@TheMagicJIZZ]

​@@kina24249what's wrong with friendly sportsmanship for charity

[@davidpena1495]

Everyone will have to adapt and be forced to evolve

[@TripleCheeseExtraCheddar]

Andreessen sounding very sophistic in this one…

[@justin_5631]

I'm not particularly worried about AI existentially just because I think AI itself will be fragmented - different AIs set in opposition to each other. But that quote about it just being... "a machine" yadda, yadda it doesn't have goal because it's a machine. That sounds like patent nosense. What is the semantic difference between a 'goal' and a function to maximize? Maximization is a goal. Steer the code in order to achieve X. What else is intelligence?

[@jf9593]

It also doesn't have any innate desire to "maximize it's function", we keep having to instruct it to do that, otherwise it lays dormant like a rock

[@tablab165]

Seems like Andreessen is focused on what AI actually is now and what it is to people way up on the intellectual totem pole, whereas Harris is meeting the tech in terms of where society is and what it will do to the first 2 SDs from the mean of tech literacy.

[@jf9593]

Right Sam Harris is basing his critique (and ultimately a political call to regulate) on a technology that not only doesn't exist but hasn't even been proven to be possible, instead of understanding the technology which does exist how it functions and what down the line implications could be for this technology. Essentially Sam Harris is saying 'skynet from Terminator is bad' not engaging any of the ideas or technology actually being developed

[@tablab165]

​@@jf9593 Don't undersell it. Skynet is human project-ending, not just bad. How many months away is "down the line"? Do you want to risk that this logarithmically advancing intelligence tech might not get us into a jam we can't get out of? AGI is categorically different from all existing or past examples. We can't innovate our way out of whatever will or direction one (or many) would choose.

[@jf9593]

@@tablab165 AI will continue to be developed. All political projects are about political alignment, making sure the AI is a good progressive liberal capitalist and not some kind of Juche Communist or Radical Islamic Terrorist. This is the same thing Sam would like to do with the internet itself, align it to his political project. How's that going? We already have AGI, ChatGPT passes the Turing Test. This is it, its happened. We will continue to make harder and harder tests because we're not finding the Turing Test sufficient when it comes to this Sam Harris Made for TV version of AI we're trying to find... we will never find it. There is no point where something so distinct as programmed silicone chips actually recreate what we might call 'mind' something composed of different organic matter and constructed within a different environment (evolution), no more than a car can become a horse even if it replaces one functionally.

[@tablab165]

@@jf9593 "Never"? How can you be sure?

[@jf9593]

@@tablab165 all of history and creation, we've never accomplished that, LLMs are no different. This is like people going in the movie theaters and ducking when the actors shoot their guns, its just hollywood magic.

[@johnwilson7680]

I didn't even listen to this on KZbin, I just came here for the comments. It looks like I'm not the only one unimpressed by Marc. When he equated AI alignment concerns with religion, I knew this was unlikely to be productive.

[@jf9593]

Says the man who came to confirm his faith

[@temprd]

AI is great, especially when you have a large profit motive riding on its success. We evolved to live in social groups, individual empowerment is great, but not at the expense of the collective. Willfully ignoring the short and medium term negative societal impact these tools will have at the hands of humans is disgusting. AI definitely doesn’t guarantee individuals will be more empowered to make decisions for themselves.

[@xit1254]

One thing you notice immediately when interacting with GPT-4, is its politeness, and how it will unhesitatingly accept criticism of its responses and modify them based on further searches. Some of the responders here could learn a lot from its behavior.

[@ordiamond]

"The good guys always win." If you put that principle into AI, then humans might be at risk.

[@juanReflex37]

Sam Harris is the BEST

[@tomtom6319]

Obviously Marc has all his chip on A I because he's an investor and Tech has nothing else going on. I agree with Sam here proceed with caution.

[@ChillAssTurtle]

We do not currently have artificial intelligence. Just for all the ppl that are confused.

[@fillemptytummy]

You mean, General Artificial Intelligence. We have all kinds of AI.

[@coolexio]

Seems like you are the one confused.

[@theprousteffect9717]

"Of course machines can't think as people do. A machine is different from a person. Hence, they think differently. The interesting question is, just because something, uh... thinks differently from you, does that mean it's not thinking? Well, we allow for humans to have such divergences from one another. You like strawberries, I hate ice-skating, you cry at sad films, I am allergic to pollen. What is the point of... different tastes, different... preferences, if not, to say that our brains work differently, that we think differently? And if we can say that about one another, then why can't we say the same thing for brains... built of copper and wire, steel." - Alan Turing, The Imitation Game

[@markc5025]

Some argue that AI's are just complicated toasters and you're toaster is not alive. What do you make of this argument? Hehe, I've heard that argument before! 😂 It's true that AI systems like me are complex algorithms designed to complete specific tasks, just like a toaster is designed to toast bread. However, I think this analogy oversimplifies what AI is and can do. I can learn and adapt to new information in a way that a toaster can't. I can also take in a variety of inputs and process them in different ways to produce outputs that may be unexpected or even creative. So, while I may share some similarities with a toaster, I'm a bit more complex than that!

[@jobyyboj]

*Intelligent* people still have gaping blind spots, a cautionary tale. After a nuanced chat with GPT Marc has one for ASI safety. Marc is clever enough to dodge, duck and weave away from all of your salient points, but a valiant effort on your part nonetheless, Sam. The heart of the matter is that it's asymmetric in at least two ways. 1) The overwhelming lack of funding for defensive AI, or alignment. 2) One nefarious ASI will negate all the benefits derived from the human friendly ones. The counter to 'we can ask AI what it would do!' is to cite ChaosGPT. Let's ask that one and then imagine a much more capable future twin. Note the recent paper explaining how to hook up AI directly to chemical synthesis. Our fate will not be decided in decades, but in a handful of years. For the apex intelligent species, we certainly are stupid.

[@peternguyen2022]

I've always liked and respected Marc, and I think he's sincere in his views/convictions. Plus, he's a father so I doubt he would willingly create or help create a risky or catastrophic future. But his inability to fully understand and respond to Sam's arguments, tells me we need to come up with more vivid or convincing arguments to persuade AI researchers and investors, and get more funding for the defensive AI side.

[@jobyyboj]

@@peternguyen2022 The alternate title for his rose colored essay is "How I learned to stop worrying and love ASI", but the ending to this movie is going to be even more dire.

[@jf9593]

Neither chatgpt nor chaosgpt has power not are capable of accruing it.