Yannic Kilcher Interview (Part 1)

Get 25% off Blinkist Annual premium! Start your 7-day free trial by clicking here: blinkist.com/machinelearningst...
This is the first part of the Yannic Kilcher interview we posted recently!
LAION petition mentioned:
www.openpetition.eu/petition/...
Yannic's channel:
‪@YannicKilcher‬
Tim Scarfe and Yannic Kilcher engaged in a wide-ranging discussion on the rise of powerful language models like ChatGPT and GPT-4 and their implications. While they believe these models will significantly transform industries and jobs, they do not foresee mass unemployment or existential risks from superintelligence.
Scarfe and Kilcher think most people will adapt to work with AI, as they did with smartphones. The key will be learning to effectively apply these new tools. Regarding concerns about bias and misinformation, they acknowledge these are issues but not fundamentally new problems - people have always spread misinformation, and we must develop better digital literacy and detection methods. Regulating the technology itself likely will not solve underlying social challenges.
Though skeptical of superintelligence, Scarfe and Kilcher expect models will become very capable in narrow domains, sometimes in ways beyond human understanding. But human judgment will remain critical to properly apply and constrain them. They believe progress should be encouraged if we are thoughtful and deliberate. Overall benefits likely far outweigh risks.
Rather than limiting access, Scarfe and Kilcher support distributing models widely so researchers and users can better understand, improve and constrain them. Regulation and oversight still matter, but restricting progress seems misguided and unlikely to succeed.
While powerful models may transform our lives, Scarfe and Kilcher expect human creativity and values to endure. With prudent development and policymaking, language models could usher in a new era of possibility. But we must be vigilant and deliberate to ensure the responsible and equitable progress of this technology.
Overall, Scarfe and Kilcher express optimistic realism about the rise of language models. With open and proactively guided progress, they foresee emerging capabilities enhancing rather than diminishing human potential. But we must be willing partners in developing and applying these tools, not hapless bystanders. In shaping our future with AI, our values and judgment will be vital - these technologies remain a means, not an end. Their promise depends on human wisdom and oversight.
TOC
0:00:00 - Introduction
0:03:09 - Background and Motivation
0:03:29 - AI Frameworks and Approaches
0:06:05 - Impact of Language Models
0:18:54 - Misinformation and Autonomy
0:29:43 - Superintelligence and AI Risk
0:45:15 - Discussion on the fear of superintelligence
0:51:37 - Understanding GPT-4 and its capabilities
0:55:55 - Comparing AI systems to human intelligence

[@MachineLearningStreetTalk]

Get 25% off Blinkist Annual premium! Start your 7-day free trial by clicking here: blinkist.com/machinelearningstreettalk

[@clarkd1955]

Yannic is excited about AI and it is unbelievably contagious. This channel and Yannic are extremely coherent so please don’t change your content. Tim, you are the most logical and mentally organized of any of the people I have ever followed. Please don’t change anything and absolutely let the rest of us learn up rather than you watering down your content. Love your channel. Thanks.

[@plafar7887]

I disagree that MLST content is incoherent. It's deep, whether you agree with guests or not. I like it.

[@maxziebell4013]

I just know him with his glasses… so charming in person. I'd love to see more of him like this!

[@Murphyalex]

Yannic's ability to constantly ground the discussion in examples and contexts of current and prior generations, mixed in with how this affects our experience with technology and gets at the differences that hide under the surface is really impressive. For example, pointing out the older generation implicitly trusting of text in emails, which my generation (same as Yannic's) would find silly, but we were the same with images, but now the younger generation doesn't have any sense of implicit truth from either. I had never heard it described in such a clear way before. The same with the effect of smartphones on pub conversations. I just about remember pub life prior to smartphones and recognise the dramatic change. You often have some great guests on your show, but honestly, ones like these often hit the hardest (in a good way).

[@marcolerena456]

Everything Yannic said about consciousness, language, humans and AI differences was so spot on, and exactly the arguments I've been making since this whole thing started. Its all a word/definition game.

[@AICoffeeBreak]

Cool! 😎😎 I'm so curious about something: Did a language model assist you in writing the video description?

[@MachineLearningStreetTalk]

Yep, anthropic did the description, GPT script did the TOC!

[@Soul-rr3us]

Yannic is awesome. Such a fun and nice guy.

[@verakalinichenko719]

Tim, you contend is great, I love it, I learn from it. I love that it is not easy all the time, I use it to discover, you cannot please everyone. Ty for doing this!

[@pennyjohnston8526]

A magical duo, hope both startups are doing well - never reach the point when your opinions are constrained by ''Marketing'. Now needing a computing role for 'when an engineer uses poking biology processes'. Grounded and refreshing conversation - Thank you again !

[@NelsLindahl]

I enjoyed this chat. Thank you for making the MLST series of content.

[@LeonvanBokhorst]

🙏 what a valuable stream of amazing content street talk provides. Thank you for the effort. Much appreciated 😊

[@MarcelBlattner]

Very nice interview. Down to earth and honest.

[@AliMoeeny]

this is amazing, thank you Tim and Yannic

[@snarkyboojum]

Very good discussion!!! Thank you.

[@diga4696]

The emergence of an informational economy, as underscored by the rise of AI, calls for a profound reshaping of our world. In this new era, information is abundant, and its value is amplified by sharing, not hoarding. This principle is beautifully embodied in Yannic's work. His efforts to democratize knowledge, to share insights and breakthroughs freely, are a testament to the power of collaboration over competition. This shift towards an informational economy is not just a societal transformation, but a deeply personal one. It challenges us to rethink our roles, our values, and our interactions. It's a call to action for each of us to become active participants in this new economy, to share our knowledge, to collaborate, and to contribute to the collective wisdom. In this consequentialist reformation, we are not merely observers but active contributors. The potential rewards of this transformation are immense, promising a world that thrives on innovation, equity, and resilience. And in this world, we are not just consumers of information, but its creators and curators, shaping our shared future through the knowledge we generate and disseminate. GPT4 is only a beginning.

[@CandidDate]

If a chatbot wrote that I'd be a bit worried!

[@PazLeBon]

it can still go either way tho. 'Traditionally I've believed in the 'good of people and have many socially orientated ideals. ID say I was influenced by people like Jaques Fresco. But more than 50% of people around me now think differently to me, its more about self and mine than society for so many of them. my family, my kids. That, aligned with inept and corrupt governments and greedy sick corporations, things things havent really become more social or empathic or given us more freedoms, people still starve and they are killing the planet. I don't think I beieve anymore that we will get the ideals half of us want.

[@missshroom5512]

Love the podcast 🌎☀️💙

[@shunmax]

Well your content is not always easy but it's for sure the most fascinating MML KZbin show.

[@vertonical]

Thanks for the excellent episode as always! (btw where can we buy the mug? :D)

[@commetking9746]

THE absolute best content channels on I've ever seen its just so magical exiting and inspiring every episode! So please please don't change a thing 🤩👍🤟although perhaps Tim could sport an eyepatch 😜

[@snow8725]

I just gotta say a thing, as an Artist... I don't see AI art as something to be afraid of, actually, it has made the process of doing art work so much more painless! I don't have to put myself through pain to get a finished piece just right, the amount of time I would spend trying to perfect a nearly finished piece to really bring out the quality and clean up all the mistakes is 10X the amount of time, probably more, that I would spend to get to that point. For every mistake I fix, sometimes I can make two more mistakes in its place and it is just so... pointless to struggle so hard with that! Well, now, I don't have to worry! Because image to image generation... Is like an absolute godsend! Is it going to make me quit being an Artist? HELL no, I do MORE art now because of it! Because it can fix my mistakes for me. Through the right combination between visual input and text based prompts I can take my work to new heights! The only issue I have with AI art is... Not its existence. Its existence is very welcome and very celebrated. The issue is whether or not the model has been trained on the work of artists who did not consent to having their work scraped and used to train AI. That is what makes me very uncomfortable. I'd like to see ways of verifying whether or not a model has been trained on artwork that does not violate anyone's intellectual property.

[@SimonWilliams0]

Great discussion

[@norabelrose198]

Yannic should take off his sunglasses more often, it makes him look more approachable and relatable

[@mattizzle81]

Glasses only come off when it's getting serious.

[@LukePluto]

i think that's why he wears them

[@paxdriver]

I always assumed it was a lazy eye or something 😜

[@milkenjoyer14]

I think he started wearing them in the beginning because he was afraid of being deepfaked 😆

[@electric7309]

​@@milkenjoyer14good one!

[@rct999]

I remember this guy was the only one who actually explained what the hell a 'moat' is 😁

[@Murphyalex]

I wondered about this when everyone started talking about moats. For us native English speakers, we all know it as the water around a castle that helps to defend it from attackers and the metaphor in terms of this mental picture is really unique and colourful. Then I realised probably not a lot of L2 speakers have ever heard the word before and it must be unusual 😅

[@MatthewKowalskiLuminosity]

Here in Michigan. The Radioactive Boyscout. This research would be excellent off planet. Lots to do with terraforming, star formation stimulation and more. Excited about further open models and tools for guiding them.

[@klammer75]

Yannic is the best! I’ve never been more aligned with how another things about this subject than him….cant wait for the part 2 of the interview!🥳🦾😎

[@live_first]

In response to: Yannic at "56:13 The real world is different first of all it's exponentially bigger". That is the same argument used comparing Chess and Go. When Full Self Driving in the rela world becomes accepted then what argument are you going to use ? I can see that being a possibility in the next couple of years.

[@TiranoBorgeano]

the discussion about rethinking and shifts with text and photos media at 21:30 reminds me of Marshall McLuhan on critical media theory, you should interview his grandson Andrew McLuhan to unpack that sort of discussion more..it might appeal to many of your viewers who say that your content is too abstract

[@rick-kv1gl]

love it.

[@appletree6741]

Some AI researchers (who are normally support open software) argue you shouldn’t open source LLM because of the associated dangers. We never open sourced nuclear weapons and even if we did nuclear weapons need vastly more resources to build than LLMs

[@vallab19]

I totally agree with Dr. Yannic Kilcher's AI position, its an eye opener. The progress in AI makes us to realize that there is nothing so special about human intelligence, therefore intelligence machines are able to perform all humans jobs that humans can do. According to 'Zero Work Theory' (1980) the progress of AI along with other scientific innovations make the HUMAN LABOUR FOR WAGES obsolete, sooner than later.

[@PazLeBon]

i find that a bit nonsense tbh, there a hundred types of emotional imtelligence and it will likely never have one of them . its still iike saying a computor is more intelligent but it clearly is not

[@vallab19]

@@PazLeBon I could not make out what you are pointing at on my comment. However I made a minor editing in my original comment.

[@RonponVideos]

It’s kind of odd that he describes the doomer position as a religion, while assuming his position of “ehhhh I’m sure it’ll be fine, I don’t even know what these people are talking about because I stopped reading,” isn’t.

[@pretzelboi64]

The AI doomers are almost always people who couldn't even write a line of code to save their lives. It's a little hard to take them seriously

[@swayson5208]

The underlying factor around alignment is Game Theory.

[@TiranoBorgeano]

this is why Categorical Cybernetics uses compositional game theory as one of its presuppositions

[@swayson5208]

@@TiranoBorgeano Oh cool, haven't come across that domain before. Definitely curious to learn more. Any good resources as a primer?

[@opusdei1151]

Thats a good one :)

[@singularityintheround]

Thank you for an inspiring and useful discussion! ...IMHO a little too dismissive of the potential for AI to accelerate and exceed our ability to predict negative outcomes. We should not base our predictions on past models when dealing with an unprecedented technology. I wish there was a little more acknowledgement of the uniqueness of the moment

[@PazLeBon]

but its still built on knowledgge that we created and we can find that info without these tools, its just that it will typicallly take a lot longer :)

[@someonespotatohmm9513]

@@PazLeBon Yea but it takes longer. If the AI rebels it is not very helpful that we will catch up in 10 years if it took it 1 hour to make the same improvement. That seems to be the main problem the safety ppl point to when talking about the existential risks of AI. All our past technology required a human in the loop. We could always stop progres and there have been 2 examples (nuclear and biological weapons) where we said no this is to dangerous, we won't research this. AI is unique in that it can improve itself without ppl in the loop. And safety mechanisms that reintroduce the human can be worked around by the AI.

[@maxziebell4013]

Competence is also a great word (Dennett) 18:00

[@maxziebell4013]

When is part 2 coming?

[@MachineLearningStreetTalk]

Not for another month or so, sorry

[@fotoyartefotoyarte1044]

What?! I Thought the sun glasses were a permanent part of his phisiology =o

[@michaelsbeverly]

_they do not foresee mass unemployment or existential risks from superintelligence._ What is the defintion of "mass" in this prediction? In 2007-2009 unemployment went from roughly ~5 to ~10% and the effects were "massive." I guess if you don't lose your job, it's easy to be dismissive. In the great depression, the US employment was in the ~25% range, not many of us alive today in the first world can imagine what that was like, certainly it was a "massive" event. No? When Henry Ford started building Fords it started a revolution in transportation. But it wasn't overnight or even in one or two years, it took time for average people to give up horses and buy a car. There was a shift, leather workers, blacksmiths, carriage operators, etc., eventually were unemployed, but at the same time, the auto industry created massive numbers of new jobs. AI replacing jobs in industry A means industry A workers need to find new jobs, but, unlike the blacksmith who could become an auto mechanic, industry A workers find themselves in a market where industry B workers and industry C works and so on are all also looking for jobs...the difference between AI and the shift in other tech jumps is how much across the board the AI enhancements work -- who's off the hook here? Janitors? Plumbers? Ditch diggers? I think depending on how you define "mass unemployement" this prediction is wildy inaccurate, the unemployement will be massive if the word "massive" means what I think it means. If we can agree and say 2007 was a massive hit to the economy and it destroyed a lot of things, isn't that fair? then, yeah, if yes, and that was "only" a 5% shift in unemployement, aren't we in for a world of hurt? Isn't it likely that in the next 24 months 5% of jobs will be lost and that it'll be across so many domains that it's not like a blacksmith who can become a mechanic but more like a guy in the American Great Depression who simply cannot find work because not enough jobs exist? And is it possible that 5% is optimistic? As to the "foresee" as it relates to existential risks, this seems to be an awkward and possibly grammatically inncorrect sentence. Not "foreseeing" a risk means you believe it's mythical and doesn't exist at all. Like, "I don't foresee Zeus destroying humanity," which is a sentence that only makes sense if it's being said sarcastically. "I don't foresee Putin launching nukes," is much different than saying, "Putin would absolutely never for any reason launch nukes, in fact, it's a joke to even talk about it." It would be cool to have the language used here a bit more concise, although, when I hear the accusation of us that aren't so optimistic (or Polyannish in my opinion) are in a "religious cult" and that we should be laughed at and mocked I sort of get it. Why even invite Connor Leahy on the show Tim? If you think he's on par with a flat earther, isn't it a waste of time to even talk to him? Or is it just the entertainment value? Or, if you do believe there is existential risk, you can't say you don't "foresee existental risk" as a way of saying you believe it's minimal. These are two different things. If there is existental risk, like say Max Tegmark and Judea Pearl aren't insane crazy cultists that don't have more brain power than a flat earther and a YEC put together, then PLEASE PLEASE stop the mocking, the laughing, the use of derogitory and nasty language to describe those of us that believe there is a bigger risk here than you do. Please. As a side bar, I was raised an evangelical young earth creationist, so I know first hand what's it's like to be in a brainwashed cult. I get it. Maybe Max Tegmark is an idiot or brainwashed or dumb or has an agenda to scare people for some reason, but it's hard for me to believe that. Really hard, because if I have to put my interpretation of Max and Judea Pearl and Stuart Russell and Robert Miles, and yeah, even Eliezer and Connor on the SAME footing as I did the YEC creationist "professor" from a Texas Bible School (who sounded so reasonable) then I'm screwed. I have no way to correctly to guide myself. I might as well become a nihilist and give up. Because I was once in a brainwashed cult and got out (by a stroke of luck, ironically the 2007 crash helped in this, silver linings) I do understand what it's like to really really really really believe and "know" something that is totally insanely wrong. I get that and I get the power it takes to change one's mind. I was near suicidal leaving Christianity, it's a mind virus that is very powerful. So I get both sides of this debate are just digging in....but seriously, am I really believing cultish naive idiots AGAIN? I might just have to get out a rope. Jesus.... So far, and I've listened to a lot of talks, debates, etc., the people who are the techno-optimists seem to have the same argument as this: "those guys are dumb. Men will never achieve heavier than air flight, it's impossible to fly like a bird."

[@ViktorFerenczi]

When is Part 2?

[@MachineLearningStreetTalk]

we published part two first. It was the one from about six weeks ago. I have just been filming more stuff with Yannic this weekend though so watch out for that.

[@AsgerAlstrupPalm]

What a hero

[@BrutFab]

in one way its exciting for the artists who now have access to the new tools to express their art and it its a great thing for all ppl to be able to use these tools. but on the other hand most artists spent years honing in their particular skills . for example as an illustrator, for them inevitably theres a sense of loss, because suddenly what they spent time on is not as valuable as much. Yes you could argue that art is about deeper expression not about particular median and now they can just learn new things, but time is an unrenewable resourceand. Also, to an extend, their art practice might have emerged exactly because of the countless hours of practice. are the time they spent honing in their skills partially wasted? And yes, you can say this happens in history, like the invention of photography to painting. The thing is the speed of change happens so fast now that no one really know how to react. It will be a question that every artist need to graple with. just imo

[@PazLeBon]

comes back to a basic again. skills were not realllllly ever meant to e monetised, they were meant to create and improve society. id have every business non profit :)

[@aldousd666]

I think of all the people in the field that I know right now, largely because of this podcast and lex, Yannic is the guy I identify with most. The stuff he said about knowing your audience, the theory about humans being less significant than supposed, and even the damn sunglasses fixation, I'm there, man. And I know the CS theory. I know the basic neurology, I know some of the other relevant theory. I just think like Yannic, in as much as I can tell from what he's chosen to publish about himself.

[@paxdriver]

"Yannic Kilcher? Excellent."

[@Low_commotion]

I'm getting that mug

[@PazLeBon]

It's kinda just an upgraded internet in many ways , one thats essentially ten times quicker. The knowledge itself is all ours and all out there of course ;)

[@sofia.eris.bauhaus]

it's so strange seeing him without sunglasses. seeing him, my brain somehow crossed the wires with Andreas Kling, wondering what he has to do with machine learning and why does he look so strange 😅.

[@paxdriver]

I think it was physicist Michio Kaku who built a particle accelerator which could produce enriched plutonium or uranium or something lol his parents let him build it in their garage while he was still in high school.

[@CandidDate]

Right up to the minute, we should keep alert as to what people with the "wrong" incentives will be able to do with this "tool."

[@PazLeBon]

having used it a lot for over 6 months. id say it can create spam, fix a few bugs in code, read stuff and then forget it 2 mins later :) i find it annoying half the time. kinda stupid when its suppsed to be bright ha-ha

[@Rockyzach88]

Pirate Bay AI shall never die!

[@6_nikki_9]

LETSSS FUCKING GOOOOO

[@ethansmith7608]

oh my god he has eyes

[@uselessrobotics5383]

He's right, a lof of people like your work. :)

[@tostupidforname]

lets go

[@AP-dc1ks]

Should have added more green light :'(

[@cruvadom]

Didn't know he has eyes

[@_ARCATEC_]

💓

[@DRKSTRN]

Be mindful when People say in the Abstract. Abstraction is Black Boxed in Commercial and in Conceptual is it just Writing something down. The Black Box the Process as to how you Work with that Abstracted Information. Leaning on there existing Methodologies and Practices that go beyond "Abstract."

[@DRKSTRN]

For myself the Logic is Correlated to Logical Movement. How do I not Stub my Toe. Logically don't Place my Movement in a Place to Do So. At What point is Informed Movement also Statistically working in Unison as we go back to the Neocortex? Switching Mode towards Automatic. Versus Course adjustment in Movement. Does that have to be Thought Out Loud? Or does that Logical Movement Happen? Driving may be the Better Example Here. Care Coming at you, And Where in Your Line of Vision in that Instant without Thinking does Logic Inform the Way out of That Situation? As Statistically that Situation doesn't Happen Enough to Be Statistical given Cost of Crashing in the Equation. Influence there Sure, but this is Conceptual and not some Binary Commercial Thought.

[@DRKSTRN]

Testable Reality is a Great Thing

[@DRKSTRN]

New Forms of Art and that's the Point

[@DRKSTRN]

Value is Content.There is Another way to Say This As Well There is Already Too Much Content to Watch Ever Unless you Are Picky. Boredom is a Powerful Thing

[@DRKSTRN]

I would Argue Individualization is Key. And something that is wholly Lacking in Our Understanding of the Complexities of Existing as Such as in the Hermit in the Cave. Plus there is a Really Nasty Angle to this in Terms of Survivability of a Certain on Average Higher Intelligence Population. Half the Life Expectancy for the Sake of Intelligence is a Funny Thing. When you Compare that to other Troubles. Problem is Thought Bubbles and Discomfort that is Inherit in Consumerism that is Unavoidable. People just want easy information and it is a really dumb equation after that. But also at the Same time Commercial has its Place. This One is Just Arguing for the Same Conceptual thought that Creates Commercial Thought. That's it.

[@vlogsofanundergrad2034]

Wait, why does Yannic look so different here? Do the goggles make such a huge difference? Unbelievable!

[@ikotsus2448]

"Not a super God like (superintelligence) because In some point there is a limiting factor, if quantum physics, if it is the availability of hardware, if it is the availability of power" At what point? So this is our safety net? -The adult human brain runs continuously, whether awake or sleeping, on only about 12 watts of power (Princeton University Press)

[@JD-zz]

Heyyy I know that voice.

[@wolfRAMM]

stop making deep fakes of Yannic if you can't generate sunglasses!

[@Rockyzach88]

Ironically I think a "left wing gpt" and "right wing gpt" could be a good thing because unfortunately a lot of divisiveness caused in politics is due to ambiguity and the exploitation of ambiguity. What destroys ambiguity? These chatbots! Good luck getting either one to say that right wing politicians support worker's rights policies lol.

[@neurojitsu]

I'm enjoying the content on this channel, but it seems to me this conversation has two important perspectives totally missing from the conversation: humility and total lack of concern for societal disruption's very real impacts. When you discuss the impact on autonomy and creativity and societal progress, the discussion is a dissociated one from the everyday reality of your fellow human beings. It lacks the quality of consciousness that you so eloquently talk about on this channel as an experiential phenomenon. The analytical (in the Henri Bergson analysis versus intuition sense) take here strikes me as rather two-dimensional, in a school debating society sort of way. The disruption to careers and ways of earning a living is not an abstract problem, solvable via analytical reasoning. Or rather, it IS for you, but not for those whose careers are disrupted and who might not possess the learning agility and 'agile' work cultures of circles that you walk in. People whose living is disrupted by AI will have to retrain, and labour markets need time to adjust and reprice the eventually re-designed jobs and find new market equilibria through the imperfect forces of supply and demand - largely without protection from regulation or government subsidised retraining. People have rent and mortgages to pay, so is it really so surprising that people resist your technological envangelism and excitement? Perhaps they are not scared of the technology, but of losing the roof over their heads? The real issue with AI is the SPEED of disruptive change. Yannic might "always want to make the best case" for a new technology, but that is a priveleged - some might say arrogant - stance that he takes that ignores the complex systemic forces that people "subject-to" (to use a Keganism) forces outside their control are sometimes unable to respond to with "agency". You will probably think I'm getting political, and thus off-topic. But I am merely trying to make the point that those in the AI industry should not be surprised when other stakeholders in society view their innovations as posing a risk if the AI community takes no responsibility for - not even to think about - the impact shocks of technological change. Some personal context may help further my point. I have a background in organisational learning and culture change, at executive levels in various international companies from FMCG to telecoms. As an HR executive and organisational development practitioner, I have led the design of 'change programmes' encompassing soft skills and behavioural change dimensions of "shocks" to an organisation's set of core capabilities, caused by things like competitive pressures or technological opportunities or labour market shortages. In a large organisation of tens of thousands, its culture and capability change takes years not months. Now imagine scaling that to a societal level, and we are talking decades. This is what I mean by two dimensional thinking about the 'impacts' of AI. You might convince an artist that she can theoretically adapt to new technology and create art in a new way, but when she wakes up the next morning she has a reputation and a following to protect, that took her years to build up. The labour market for AI, data science, and other related computing skill sets and intellectual abilities is a much faster-moving one that that for artists. This is why Hollywood is striking right now, not because they are Phillistines but because they have the human right to challenge who are you (ie the collective AI community, and particularly the VC-funded startups and big tech platforms) to decide my job will no longer be relevant? Try to empathise, by putting yourself in an artist's shoes, for a moment and feel (NOT THINK) what that means. Some humility and demonstrating care and understanding would go a long way towards passifying such fearful voices, but discussions like this do not. I collaborated once with a speech writer for Tony Blair. He said to me that one thing he teaches all leaders (including many CEOs he works with) is that when you talk to one person in your company, imagine you are talking to everyone. Because that one person will talk to another, and so on. Yannic would do well to reflect on this: you are not talking on KZbin just to the AI community, you are representing them in the wider community of interested stakeholders.

[@MachineLearningStreetTalk]

This is a wonderful take. Thank you so much for taking the time to write it. I would need to think about this more, but it seems to reveal a deep political schism in thinking. Which is to say should your job have intrinsic value and should you have protections against your job being automated as a matter of principle. The former Soviet union is a great example to consider here where jobs were essentially created with no particular purpose and conferred no particular value. I agree that we need to be conscientious and sympathetic to the plight of people who are losing their jobs. I also can accept that there is a degree of arrogance from the tech folks who feel that their skills are not so much tech skills, but rather an intellectual "flexibility" to take on any new task; should the circumstances demand. "if only those plebs would adapt with the times." - I will think more on this and apologies for any grammatical or presentation errors. I’ve just dictated this on my mobile phone.

[@neurojitsu]

@@MachineLearningStreetTalk Thanks, yes I agree it may be something that deserves more representation in the thinking of AI leaders, which was my reason for raising it with you as I felt - correctly! - that you would care. Happy to exchange ideas...

[@neurojitsu]

@@MachineLearningStreetTalk I just had a further thought that connects to the speed of change point, which I thought I'd just share. BTW I touch type at about 90wpm, an old-school skill, so I can shoot off long comments fast... not always a good thing! There's a Deloittes report from 2018 I think that looked at the link between productivity and technology, that sought to examine the question of why economic productivity had stood stood still for some 10 years plus since the 2008 crash, when so much new technology had been implemented in that period. Their conclusion, in two paraphrased words/metaphors, was overwhelm and indigestion. On overwhelm, things like email and mobiles had just added more communication on top of people's workloads, without taking anything away. It is my observation, that few companies' leadership teams seek to simplify the project workload of companies when they set off in new directions: every new strategic imperative or project is an add-on, nothing is taken away. The report blamed this poor prioritisation for eating up any of the potential benefit email and mobiles might have otherwise delivered, neutering the potential efficiency gains. Moreover, they said people were never properly trained how to integrate the new technology into how they worked, a common IT implementation failure in the literature: companies' failure to embed new ways of thinking and acting, means that people just find workarounds to work in the 'old way' using the new technology! Maybe AI is a different order of change altogether, but these warnings from history are nevertheless pertinent IMO. On your point about the intrinsic value of jobs and protection, I'm not sure that I'm advocating for this - though I do believe every human being has the right to dignity, self respect and a living wage, all things that currently people seek from their work. In the absence of some major social reengineering, such as universal basic income, then any job destruction is going to have major impacts on not just financial outcomes but mental health, and all the associated costs to society of bearing that burden given our depleted NHS. So at a policy level, automation's impact and economic costs need to be borne somewhere and the question is who will pick up this cost and manage the necessary retraining and transitional challenges? Regulation could be part of the answer, but our current government is showing no interest. I've seen taxation suggested as a potential answer, ie taxation on the big tech firms that seeks to redistribute wealth from the firms causing the disruption to where the impacts are... but I see that as a very complex thing, and anyway government taxation is not ringfenced, and the evidence of governmental overspends on IT and procurement projects gives me little confidence in a government-managed solution - it's just too complex and unwieldy. Education seems the only answer, with a much higher investment and far fewer controls. And don't get me started on universal credit's punitive approach to getting people back to work: that has to change, or we will have mass misery should the worst case fears of AI job armageddon become true (I don't have a view on the likelihood, but I share the view that we are totally unprepared).

[@jonathanedwardgibson]

Yawn. I like Yannic, but he keeps archaic comically-simple ideas of mind and language - of course he’s satisfied with this progress. Riddle-me-this: when you flip an Ai off+on why doesn’t it gush to you about near-death-experiences? One wee hour my son woke me up complaining his grandfather was bothering him & wouldn’t let him sleep. Blinking from the light turned on, the phone rings and ex-wife explains her father has just died a few towns over. i said, “We know.” If that cruel, petty, mean man can cross over and pester my son from, beyond death, I’m paying attention. It’s not that Ai will be super-monster, but what humans do to each other with Ai that should concern us.

[@swayson5208]

Depends on the audience. He speaks to folks who are not experts at philosophy and alike, and value this simplicity as a seed to further learning and insight

[@annawilkin9067]

He did seem a little green at some points

[@Soul-rr3us]

I saw the part about the green lighting, so i guess you are joking. If not then no way. He has a a PhD in this, published research, and has been following ML research for years.

[@annawilkin9067]

@@Soul-rr3us Oh well then, if he has a PhD then nothing to worry about.

[@Sawa137]

He's a well meaning, nice guy, but he refuses to think about the big picture AI safety problems, too absorbed in his own work, typical sadly.

[@linkawaken]

An AI practitioner's position on AI safety is commonly like this, as they are fairly adept at and engrossed in their work, but have not yet familiarized themselves with the arguments, seeing no need to (since they are already practicing AI, they see less need to). Introspecting on my mental model of what someone's position on AI safety problems will be, I see many parallels with politics an religion at play, where one only needs to know a small amount to predict their positions with high accuracy. *Correlated thinking* is when otherwise-unrelated beliefs like positions on gun control and climate change are associated with group membership, and the same thing is at play in AI safety debate and discussion, I suspect. Within the broader AI safety community are sects; the one most concerned with bias is the tech community, due to its political orientation to the left. Increased concern for bias is inversely correlated with concern for X-risk (note the focus on bias in this conversation). Stepping back to the overall camp one falls into (extreme risk from AI is to be taken seriously, or not), perhaps there is a strong psychological defense mechanism at play to tell a story of progress toward good ends in perusing one's life work and passion.

[@lolitaras22]

Just an algorithm manipulation comment.