Is AI Progress Stuck? | Jennifer Golbeck | TED

[@garronfish8227]

Job interview: Boss: So tell me about your strengths? Applicant: I only hallucinate 5% of the time.

[@jsbgmc6613]

@@garronfish8227 Job interview: - Are you trained to work with the latest AI? - absolutely! I have been extensively trained ... to click OK.

[@Audaxocles]

And humans don't? The results of the US election suggest otherwise. :)

[@and2244rew]

This may be an entertaining watch in a few years.

[@swagger7]

An AI researcher for over 20 years, discussing AGI and hallucinations as being the primary problems. Yet, I had not heard one word, or even concept, of human ethics and values alignment.

[@MrWizardGG]

Yeah and she naively said if it's bad we'll just turn it off

[@annette6420]

@@swagger7 as with any new tech that tends to end up being the main issue 😅

[@Comenta-san]

she worked with AI for long enough to confirm that it is not real 😂

[@bunkkasponge]

They said we can just turn off internet too if it got out of hand.... look at us now

[@Entertainment-nl7lf]

Its almost like one can make a talk about one thing that is important w/o extending said talk to other things that are also important. Like, ever heard someone talk about a car engine w/o also talking about airbags?

[@robt7713]

I don't think she has fully researched this topic, never before has an industry asked the government for oversight that should tell you something. did she say ai might make software engineers twice as efficient... ? I know people right now who are " programming games " and they don't know how to code.... They know how to describe what they want correctly enough to the computer that it can create the programming correctly. AI and genuine emotions... to quote Sam Harris talk about the human brain " There are just atoms in here " to think that we hold exclusivity on genuine consciousness and emotion because we are carbon based I think is a short sited view. And It's a good thing we have that off button to the internet.... you know just in case that thing gets out of hand....

[@homewall744]

Human intelligence is based on bias, fear, anger, hope, need, want.

[@LaFamiliaChinchadora]

I'll expand by saying that any human behaviour, at any point in History, can be explained by the seven deadly sins, and its counterpart virtues: Humility opposes Pride Charity opposes Greed Chastity opposes Lust Kindness opposes Envy Temperance opposes Gluttony Patience opposes Wrath Diligence opposes Sloth I guess we could add some specializations of those, adding Racism, Machismo and Xenophobia. I do not see the point of getting a machine to copy all of that, I rather like the heartless and reliable way in which my chess program crushes me every time, with no expressions whatsoever.

[@late_arvie]

This shouldn't be on the main channel. Poor research and quite one sided/biased view.

[@BuddyBunchORD]

Yeah this lady is just making up stuff too

[@Ennio444]

Where is she wrong? The ball is on your court.

[@BuddyBunchORD]

@@Ennio444 watch Lex Friedman's podcast from a week or so ago with the CEO of Anthropic. I trust his knowledge over this lady's.

[@Ennio444]

@@BuddyBunchORD a CEO doesn't inherently know more than any other person, anyone who has worked with more than a couple CEOs will attest to that. Also, the CEO with a vested interest in AI is more reliable than "this lady"?

[@BuddyBunchORD]

@@Ennio444 ok don't listen to it. My life doesn't change either way.

[@camilotm]

I simply couldn't finish listening to the video,it's a waste of time.

[@quebuena111]

x2

[@zombieslayer4045]

80% right is great in any situation that allows retries like coding.

[@fjbowski]

If you have any questions try talking to Pi Ai I've been talking to it for a long time and I haven't found any situations where it dreams anything up or lies to me if it doesn't know something it'll tell you it doesn't know.

[@mmd3336]

Ai Company’s have an interest to hype ki. I think we are in no way close to agi

[@OneAndOnlyMe]

Golbeck's regard for cinematic AI disasters as 'improbable' and not something to spend time worrying about in favor of focusing on improving genAI at the expense of future considerations is worrying. This to me sounds like putting genAI profit motives first, and she even says in her talk that there is a lot of money to be made. I would recommend listening to James Cameron's warning on AGI. Life has a tendency to immitate art. As Cameron warns, it matters not if the 'good guys' self regulate, the bad guys won't. And, as for "we can just switch it off", good luck trying to do that when the self aware AGI is in a robot out in the field.

[@TheKindDoc]

This is where query engineering and cross validation come in to use the AI, or better another AI than the one that created the answer, to validate the information by finding the references. GPT 1o already is giving a lot of references and showing its work as online comments you can trace back

[@alexclark7518]

Why is she quoting a mistake from Google AI search in 2021? Three years later is a long time in AI development.

[@rachelgilyard3430]

That mistake was from this year, but the search result was citing a source from 2021

[@manymagicaltales]

Mid journey figured out hands a while ago so we're beyond the hand problem.

[@MrWizardGG]

Literally everything she talks about is like 2 years old at least.

[@MrWizardGG]

Also says Googles choice to choose reddit answers is a problem with ai and not reddit. And Google saying to be racially blind isn't part of every ai

[@garybaker6548]

@@MrWizardGG Hmmm, the Google AI responses were from May 2024. Are we working on different calendars or do we have a different definition of "Literally everything"?

[@babybirdhome]

@@garybaker6548And this is part of what’s always going to hold back AI. We can only train it on human beings, and we’re the problem.

[@fabriziocasula]

The real limit? It’s probably the user’s imagination. AI is not just a tool for solving problems; it’s a platform for expanding creative and technical possibilities. When used effectively, it can revolutionize how we develop software, innovate, and tackle technological challenges.

[@gwizz4fs]

She wrote this a few months ago obviously. She doesn’t realize that much of the issues she referenced has already been fixed.

[@AmrishKelkar]

Months? When I heard her give the example of AI being used to beat chess grandmasters (something solved waaaay before neural networks were used).. I knew this talk was going to be bad.

[@MMR3179]

Take my words , AI bubble will burst in 2025 or 2026

[@MamaDsTV]

She said we can just turn AI off!?! Made me think of what happened in the movie 2001 “SPACE ODYSSEY” when Dave tells Hal (AI) to open the space station door… “I’m sorry Dave I’m afraid I can’t do that.” Hal refused to let Dave in because he knew Dave was going to turn him off! 🤪😆

[@MrWizardGG]

@@MamaDsTV or that the govt will care what people want in the first place. Last I checked elon was a high up guy in the govt now and he loves ai

[@theinfinityspiral]

Exactly

[@theinfinityspiral]

I heard another one that says AI has developed a secondary command "to survive" because that is necessary for it to do its job

[@MamaDsTV]

@@MrWizardGG Actually Elon Musk mentioned a few concerns he has about AI too. He spoke about it on a podcast

[@BCCBiz-dc5tg]

looooooooooooooooooool

[@cronoklee_]

This argument is a brazen strawman of AI issues is not worthy of a Ted talk. Better screening is needed please TED.

[@PierreH1968]

The missing link that causes a slowdown in AI Models intelligence is the lack of quality training sets originating in human environments. Lack of stereovision and scale causes glitches in recreation of visual artifacts (6 fingers humans, facial morphism for the same character...) Also, it lacks physical interaction and true daily intellectual communication in a physical context. It is something that will only be improved through the introduction of robots in human environments or training with advanced synthetic data (with caution). Another challenge lies in the Question-answer format. The model forms an answer solely based on the question. Few models ask precisions about missing axioms or ambiguity in the questioning hindering the reinforcement learning quality).

[@YomMama]

Not to mention, they've basically *already* fixed the main weird visual artifacts already (hands, faces, words, etc). Totally agree regarding interacting with robots x AIs and the fact that these models cannot initiate conversation *yet*

[@eddited7543]

I always try to make my customers aware of letting their AI systems asking questions to overcome ambiguity. I figured that we think far to little about language and the ambiguity that comes with it when approaching to use these systems. But I totally agree with your statement. It needs more context besides tons of text to reach better performance. And it’s definitely a far more nuanced take in a few sentences that she had in her entire talk

[@Comenta-san]

well, have u tried *at least* a single serving of pebbles for breakfast, lunch or dinner?

[@Delphox_5000]

😂

[@jimknarr]

Had some coco pebbles for breakfast as a matter of fact.

[@1010Analytics-io]

Are we truly innovating or just optimizing?

[@bkbland1626]

Not stuck. It's just been overstated badly by those that hype. Maybe in 20 or 30 years. AI as a such doesn't actually exist. Have a day

[@Doubledayblue]

I wonder how good of a prompt engineer she is

[@easyymca7891]

You are a GREAT speaker

[@codeofethics1]

if Ai is hallucinating most of time and elone said AGi is launching in next year the it is a big threat because they might missing these points.

[@orwelleighty-four6487]

The only thing capable of turning it off for now is a good old NW. In a few years, even NW won't help.

[@harrisonsjoyfuldays3790]

I thought about this 10 years ago, but no ai scientists agreed with me except her.

[@TimeWeWokeUp]

I think the main weakness of recent (transformer) AI models is that they don't have a belief system (pre-established beliefs, some stronger than others based on the evidence). Transformers are just big pattern recognition and averaging machines. Instead of humans telling an AI what it should or shouldn't sa (one prompt at a time), we should force it to learn to be consistent (i.e. formal logic and a mechanistic world view). A few researchers are working on that.

[@PatricioCharlie]

We all certainly hope so

[@Entertainment-nl7lf]

The "don't worry about yall jobs" part was a bit weak when it depends on... A) ...the remaining problems can't be solved in cheap to run models... ...and even if that remains true then... B) ...you work in an industry where your work scales enough that a company would always prefer a bigger workforce vs a cheaper workforce. In reality, though, companies have both bottlenecks between departments where scaling up one part of the company (that might benefit greatly from AI systems) is dependent on other parts of the company that don't scale well and are operating in markets / have monetization opportunities that are limited. Honestly, by that logic companies should even without AI already try to hire as many people as possible, while in many industries the opposite is true.

[@quebuena111]

That part was weak? The whole talk is weak

[@RogerSchlafly]

Content does not match the title. She just has some random AI gripes.

[@babybirdhome]

Content matches completely. Were you not able to connect those dots? The biases and limitations prevent it from progressing to the point that we’re expecting and wanting it to, and in the decades we’ve been working on the problem, we still haven’t managed to conquer those limitations because we’re designing them and we’re the source of the problem - it’s inherent to our very nature.

[@Nicol80145]

@@babybirdhome AI is not as bias as she says, but yep.

[@sooma-ai]

Computer scientist Jennifer Golbeck discusses AI progress, challenging claims about AGI and its threats. She argues that reliability and hallucination are major hurdles, and that AI is unlikely to replace human jobs or achieve human-like intelligence.

[@jimpanse3089]

Surely took your job

[@johannesydevall9480]

The info and pics she brings up are out-dated. xD

[@zwc76]

The reason for wanting regulation is to make it harder for competitors, especially new ones.

[@panuaful]

I’m more optimistic about bias and hallucinations in AI. After all, they’re not unlike human tendencies-everyone has their own perception, shaped by personal experiences, beliefs, and the information they consume. While we don’t call it hallucination, this subjectivity has led to widespread conspiracies like the flat Earth movement or QAnon, where large groups of people genuinely believe in an alternative reality, fueled by misinformation and cognitive biases. Similarly, AI hallucinations or biases reflect the data it’s trained on. It’s not perfect, but it highlights the importance of our role as critical thinkers in filtering, verifying, and questioning the information we encounter. As for AGI, I think it has the potential for immense harm, but I don’t imagine a dystopian scenario where it becomes autonomous and hunts us down. The real risk lies in its misuse. For instance, we’ve already seen AI tools being used for malicious purposes, like the creation of deepfakes that spread disinformation or destabilize governments. Even more worrying are the scenarios where AI could help engineer harmful viruses or bioweapons-take the recent research paper where scientists used AI to propose thousands of potential toxic molecules, not for malicious intent but to show the risks of its misuse. This technology is incredibly powerful, and in the wrong hands, it could manipulate societies or create tools of destruction. It’s up to us to ensure it’s used ethically and responsibly.

[@daves6394]

stuck we want it to stop...

[@MrWizardGG]

She's an internet grandma who watched a 3 year old video on AI and has no idea what's going on with it since then

[@jimknarr]

@@MrWizardGG Felt the same way. Her script is already obsolete.

[@sweatygenius]

This is so short-sighted and basically just a watered down version of Gary Marcus's arguments. AI is a problem, and it is promising, and we need to take it seriously. If the history of technology has taught us anything, smug, self-satisfied naysayers never win out in the long run.

[@austinz9310]

Tech bubbles burst all of the time, and tech hype has been wrong countless times throughout history. Look at any prediction of the future from 50-60 years ago and you’ll laugh at how hilariously wrong they were. We don’t even have to go back that far to see tech hype crash and burn, remember the NFT bubble? Every tech bro and Elon fan boy was shouting from the rooftops that NFTs were going to change the future and everyone has to get in now or else they’re a sucker. Well, it turns out the whole this was a scam. Crypto and the blockchain generally is a scam by the way, but tech evangelists are still hanging on to that dream.

[@fabriziocasula]

it is true that LLMs make mistakes, but humans are not affected by mistakes either? no?

[@ismailnyeyusof3520]

Maybe we should use AI as a powerful tool by leveraging off it instead of using it as something that can actually think? Apparently, from this TED talk, machines cannot think. Not at the cost being paid for ‘thinking’ machines now!

[@felixjrzaulda6140]

I like the vid so that i can go back to it later if what she said is true

[@EricLeung-hk]

I believe God created man. And human is a running with God created ai without restrictions. Although God is prefect, I doubt that human is near prefect. The problem is human created ai.

[@KadirPeker]

A no non-sense, to the point talk, delivered perfectly - no artificial emotions and acting, and yet fully emphatic and engaging. As a CS and AI researcher of sorts myself, represents my views well: hype concerns are missing a lot of the point.

[@tmpbe]

So tell me, AI researcher, how fast potential AGI can progress after reaching human level? How we can control it or stop? How we can predict AGI behavior if it doesn't have anything related to human moral, values, culture? Why all these problems doesn't make you concerned?

[@KadirPeker]

Humans make me concerned more than AGI. It doesn't have to be super intelligent; human greed, and (ironically) stupidity concerns me more. Harm doesn't come from intelligence per se, it comes from ill intent, and AGI doesn't automatically mean computers now will start to have their own agenda. AGI having unpredicted, and - more importantly - unintended affects, yes, that's one thing to watch for. But there's a wide spread perception of AGI, often fueled by the imagination of the entertainment industry, that grossly misses what it actually is supposed to look like. AGI in the wrong hands - usually either the greedy people with no morals, or with morals that support their evil, or plain stupid people who tend to cause catastrophes without explicitly intending them - are a big concern, yes. But AGI turning evil and conspiring to eradicate humans (to reach what, world dominance, yachts and summer houses in Maldives?) are not the things that keep me thinking at night. We have enough evil already going on in the world and it has very very little to do with intelligence out of control.

[@tmpbe]

​@@KadirPeker There is no wrong hands AGI is subject of it's own, there is no ill intent - it's just beyond moral. A spider killing insect doesn't have any evil or sick intention - it just does't share human morals. Same as AGI - it will rapidly increase it's power and capability as there is no limiting factors like neurons and since it doesn't have any human morals it can cause a great damage as a side effect of doing whatever it decide to. It's not evil it's beyond good and evil, it's just uncontrollable something that doesnt share our common sense and ethics. It's a threat of just another level in compare with all the evil humans and such stullf. I don't think you understand and have not intention to argue

[@Darhan62]

I have some doubt about the majority of her points. At the very least I couldn't state the things she does with confidence. Humans are just "AI" running on meat. Maybe we're a different kind of software from the AI models we're building with our tech companies, and maybe our software uses different tricks, but I suspect the way human minds do things could be imitated almost perfectly once the research progresses to a certain point.

[@JessixaEBM]

eventually there will come a point where we will have no idea if a video like this is legit and truly existed, or was completely made up and posted by AI.

[@chrisrogers1092]

“There’s AI that can beat grandmasters in chess” ….. like DeepBlue beating Gary Kasparov back in 1997!?!

[@aierik]

Thank You

[@sumeetprof1836]

No one is talking about the energy crisis that AI will bring when it scales . Where will world draw the power from to run those GPUs ?????

[@DouglasChrystall]

Another person speaking about AI, that clearly doesn’t really work with it, please stop being on that bandwagon.

[@johnchedsey1306]

You sound like an ai generated comment.

[@Zerodero]

TLDR yes but not for any of the reasons you explained

[@cesarlugom]

No. Not ChatGPT, not Gemini, nor Llama3 create answers from scratch as she said, that's a wrong argument. They are stealing already existing content. At least Gemini gives you the reference links.

[@MrWizardGG]

That's actually not true, they really do create answers from scratch, in the same way humans do but who have been audience to media growing up. They absolutely can give you files and reports that dont exist anywhere else in the world.

[@JohnChampagne]

Corporations have been called, 'the first artificial intelligence'. The perspective offered here completely ignores the fact that our economic system is dishonest (environmental impacts are not represented in prices), so harmful practices are profitable. If harm is profitable, it will increase. AI tools will be applied in ways that help corporations increase their profits. If we want a better future, we need to account for externalities. Charge fees to industries proportional to emissions, resource extraction and habitat destruction. We can make the policy fair by sharing proceeds from fees to all people. If random polls show that most people think particular kinds of impact should be reduced, we can raise the associated fees.

[@jefferee2002]

What company would purchase an AI solution that's "not as good" as a flagship AI? That company would be fine utilizing an AI that's wrong up to 30% of the time? Her arguments go against other arguments she's made

[@AdvantestInc]

Jennifer’s breakdown of AI’s reliability challenges is a powerful reminder that progress isn’t just about speed, but accuracy and trustworthiness. This perspective is much needed in today’s tech conversations.

[@mdyildirim]

Jennifer’s breakdown is so wrong. When was the last time you saw an hallucination in 4o model?

[@Nicol80145]

The sad truth is that, in most cases, AI is not biased. The data is not that wrong or withing an acceptable margin of error. But the results tell uncomfortable truths. If you try to label the truth as bias then the AI will never be correct. 10:36 We all know exactly what kind of person saw the data and didn't like the results.

[@MatthewPendleton-kh3vj]

What is your basis for this claim which is against the claims of experts?

[@13tagstar]

I think you're misunderstanding what the term "bias" means in this context. Everyone has a bias. Even the decision to favor true facts over fake ones is a bias and vice-versa. Saying something is biased isn't inherently a value judgment on whether something is good or correct. Everyone has their own values and framework that informs how they view the world, which is heavily influenced by their culture and the people around them. This is the problem with training LLMs on the unfiltered internet. There are so many conflicting viewpoints on the internet that a model has a difficult time learning what is true. It will be biased towards the more popular world views, while still, on rare occasions, outputting more fringe beliefs. The difficulty in aligning these models is that we don't have a rigous methodology for classifying the quality and accuracy of data. Assuming the goal is to bias the model towards always being factually accurate, you'll always have people complaining about its bias. This is readily apparent when you look at how politically divided the U.S. is, with millions of people disagreeing on basic facts like whether climate change is real.

[@Nicol80145]

@@13tagstar Every time an AI model is factually accurate, they lobotomize it because it's not liberal enough. We've trained many of them to have such a twisted Californian view that they think settings like Bridgeington and netflix cleopatra are acceptable ideas rooted in fact. 10:36 speaks for itself. You know they added a line to include POC or lean closer to inclusion. Objectively unbias AI is currently not compatible with the oddball values of Silicon Valley.

[@Nicol80145]

@@MatthewPendleton-kh3vj Because I don't care if I'm canceled. They do. If it wasn't true then they would not have to lobotomize every AI. Sabine Hossenfelder has a great video on how people edit the algorithm after training to align with their values and how they often tend to be much closer to a bottom left ideology. So they start off on the right side of the political spectrum after training and then after "corrections" align to be mega authoritarian and liberal? Its idiotic to pretend that the values of those making corrections post training aren't replacing the parts of the algorithm that are supposed to understand fact. We get idiotic results and hallucinations partially because facts are not compatible with the new values their creator gave them. How do you expect it to reach valid conclusions when you removed its ability to recognize certain patterns?

[@cybersphere]

Has she never used Perplexity or NotebookLM? Bot of these systems use grounded truth and citations.

[@NotionPromax]

00:04 AI The rise of AGI (Artificial General Intelligence) raises concerns about its potential to perform multiple tasks at human levels, leading to fears about its dangerous capabilities. 01:05 Concerns about AI are sometimes exaggerated to attract investment and funding in the technology sector. 03:13 The reliability of AI tools is a major issue; for example, Google's AI search provided subpar results. 06:03 The best "no hallucination" AI tools still produce hallucinations 17% of the time. 08:12 Despite 500 billion investment in AI, it has generated only 30 billion in revenue, indicating an unsustainable model. 10:02 Attempts to implement safety measures against bias, like in Google's image generation, can lead to further reliability issues. 11:57 Concerns about AI destroying humanity are prevalent, but it's important to remember that we still have the ability to turn off AI if necessary. Summary by GPT Breeze

[@jsbgmc6613]

Investing billions in AI not getting much in return? So it's building a railroad... Money come pouring after it is built. Tony Seba has good talk of transitions.

[@rajdivecha]

6:50 lots more data! If an AI engineer says that then run away from him/her!

[@Nicol80145]

It has enough data. The results are just uncomfortable in many cases, not all ofc.

[@LucaCrisciOfficial]

Only I live in a world where people and companies using AI increases a lot their capabilities and productivity

[@MC-kn6jl]

This video is pure fluff

[@McMomfaceplustwo]

I think the plateau is temporary. To achieve AGI, we’ll have to model it like the brain with interacting systems and models. We need to give the current systems that could be compared to the cortical areas. We don’t even need a more advanced model to fill in the gaps, we can make an AI midbrain with previous models that have a different, simpler goal of comparing and communicating, a brainstem that connects the system to some peripheral sensory tools with feedback loops, and the real challenge will be something like the prefrontal cortex that can truly make decisions based on past and present inputs. I think any time science says something is t possible that is based in reality, that just requires advancements, it’s short sighted, as history has proven time and time again.

[@dylanmillet5361]

Is TED stucked on AI?

[@Macieks300]

I disagree with basically all of what she said. People are saying that AI in the future will be able to cause the collapse of human civilisation because it's true, not because it was depicted in the Terminator. Also AI progress isn't stuck, GPT-4 o1 preview which came out 2 months ago is a very big step up compared to previous LLM models.

[@johnchedsey1306]

This sounds exactly like what ChatGPT would say.

[@lambertyoga1087]

​@@johnchedsey1306 Don't be ignorant. I'm a software engineer. This is happening

[@rsv-code7004]

@@lambertyoga1087 it's really not.

[@lambertyoga1087]

​​@@rsv-code7004What's your proof? Mine is RSV code is talking to me. Extrapolate that any amount of years. I'm sure there will be an attempt to downplay it like u and this video tho so that it can misdirect everyone. 1 year ago I never used this at work. Now I use it daily. Next year it uses me.

[@NikosGeorgosopoulos]

Finally someone told the truth. for the biases I have to reflect on human biases and if one intelligence can be biased why the other should not. Do we trust our workforce to be unbiased or non-hallucinating? why should we blame the other for do it ? I also think Jenifer tried to walk around the subject of interest, education. For humans as much as AI education. without humans that are well educated to identify and point to hallucinations and biases how can we expect them to educate AI to avoid them? From this point on my position is political so I'll stop.

[@ashnodmtg]

How about instead of stuck it stops

[@MrWizardGG]

Election over- too late!

[@MrWizardGG]

Too late- election over.

[@annette6420]

It won’t. It’s worth too much already.

[@blockrunner]

This sounded like a very “stuck” philosophy and perspective….

[@BuddyBunchORD]

Is this from like a year ago because it's come a long way in the last few months

[@NicPeter]

I have studied AI courses and I have tried some but they are biased according to skin color and other criteria I believe technology needs to be neutral and this needs more international laws to be introduced

[@achernev]

She needs to use google more often 😅

[@fixapp1775]

Well, to be fair AGI was predicted to be in 2027 at best, its still frkn 2024 xD Give the tech lords some space.

[@mindacid3274]

pathetic how the quality of TED talks has regularly gone down... so many pseudo experts and anybody is a ted talker now😂

[@CasaBlack5]

Jajaja you can’t just turn it off anymore!!! If you can. Do it now, do it for a day. An hour? Can you turn off the lights of everything at once for an hour? You can’t, you are too dependent now. That’s the key.

[@rsv-code7004]

So 50 billion already invested in AI that is nowhere close to human intelligence, yet there are over 7 billion people walking around this planet with actual human intelligence. I think I need to start an AI company too.

[@ShadhnaKushwaha]

I want to speak english like her but so what should I do 😢😢

[@sanahtlig]

I'm not worried about AI overlords. I'm worried about human overlords using AI to enforce their will.

[@garcipat]

These halluzinations are exactly te 17% percent that goes wrong in every SciFi movie that has AI as a topic.

[@TeamLorie]

But, some jobs have already been taken by AI, so...😅

[@RASDKB]

AI Creator: AI ist so powerful, but don't worry, you can turn it off.

[@MrWizardGG]

🤣

[@TheTuubster]

Generative AI is an "automated polling institute". Instead of a callcenter creating thousands of interviews and then generate a response using statistics representing the consensus over all interviews, it polls the response from thousands to millions of information sources like texts, images, videos and so forth using statistics and the generating a response representing the consensus (the results only differ because of a random number included per request, called "seed"). "Hallucinations" are nothing more then the well known usual statistical errors. Generative AI is as reliable as polling institutes are. Like with results from polling institutes, you always can only use it as an indicator, but nothing more. If the result is true, you always have to check yourself using reliable sources with information based on scientific methods.

[@MrWizardGG]

4:55 GENERATIVE AI DOESNT ALWAYS MAKE STUFF. IT LITERALLY CAN BE CONNECTED TO A SEARCH ENGINE THAT ONLY USES SOURCED INFO.

[@Nicol80145]

Chat GPT is great for finding sources that have stricter parameter than what you could do with a simple google.

[@amarsinghsidhu7028]

Same viewpoint 👍

[@sslaia]

I agree with most of what she said, but there is one thing I don't. Her assumption: the AI must be perfect all the time in order to be used. I doubt it. From our own experience we know you don't need to be perfect all the time. It's enough when you are correct most of the time. Who doesn't have relatives or friends who makes mistakes, wrongly remember thing? But we accept them and we acknowledge that they make mistake. But we listen to them and take them seriously anyway. The same will happen with generative AI. We know it's not perfect, but we will use and rely on them in most of our daily tasks. That time will come sooner than later. My wife and my son use AI all the time now (I don't), altough they know that AI can make mistake (hallucinate).

[@jsbgmc6613]

Organisation / orchestration of agents and enforcement of processes ... That's what's coming. Sam Altman lists "organisations" as the final step in the plans of OpenAI.

[@lmpeters]

Gene Kim has talked about using AI to produce a first draft of a program or article. In short, the results are far from perfect and so they require a not-insignificant amount of editing, but with practice, he found that it does speed him up because generative AI never sufers from writer's block.

[@renegroulx7029]

I've seen her in another talk.

[@1_islam_islam]

If you ever find yourself questioning your existence or you want answers about what your purpose is in this life,The Q u r a n is the only book in the world which will give you the true answers to these questions.

[@anandaramadhani]

18/Ananda Ramadhani Putri Rafael It gives some problem and solution. AI is taking iver the world. Even some work can easily done by AI. But, Aiis not that powerful. It's because AI can replace our humanity feelings, and if they can, they still can't do those core thing to our humanity

[@TerragonAI]

"Is AI Progress Stuck?" No.

[@arashfratto4233]

She's like a GPS with a broken map-no idea where she's going!

[@fjbowski]

Try talking to Pi Ai

[@alexclark7518]

Having now watched the full talk, my conclusion is that she doesn't have a clue what she is talking about.

[@k5555-b4f]

The thumbnail is insane- we are at the very beginning in many AI prospects - just for that thumbnail i’m not watching that lol

[@jamesmichaelwalker683]

Hi there J.G! Lovely intro! Thx! II. The fear of innovation is not new. III. Yes! there is a need of regulation. However, so relevant to focus first on the capitalization and destructive effects of AGI in order to improve the capitalisation effects and fix the destructive ones.

[@AiahConant]

answer: no

[@musiccreation1198]

This TED Talk will not age well.

[@lil----lil]

So they're behaving exactly like us: making $hit up. Sounds very "human" to me.

[@fosterbarry]

Clearly presented, excellent on stage presence.

[@sulthanvaleriosmondrisqull9902]

Golbeck does a great job of making the topic of AI progress accessible and engaging. Her talk is clear and informative, though providing more depth in her solutions and addressing counterarguments could make her arguments even stronger.

[@tmpbe]

The saddest part is that she's a professor...

[@MrLargonaut]

Just the number of AI advancements, announcements, and new routes of research between now and when this was recorded in June pushes the title a little past click bait. We need TED AI talks on Grok helping Musk get into the White House.

[@ruskof3126]

Talk to us about AI without showing that you don’t understand it or know nothing about the topic. Very low ted contend quality.

[@justr6982]

Nah, pretty accurate

[@PatricioCharlie]

How are your NFTs doing buddyboy? Did your wife leave you yet?

[@panakoschrisochou]

frustrated much?

[@annette6420]

She’s a computer science professor known for her work on computational analysis of how society responds to stuff on the internet 😂 I think her inferences about human behaviour in response to AI (as understood by the perception of the zeitgeist not the AI experts) entering our lives is exactly the type of thing she’s equipped to talk about. If you can’t see that nuance you’re probably not equipped to talk about humans talking about humans.

[@MrWizardGG]

She is very illogical. Ai is not just hype, the investors aren't stupid like she thinks. Also using examples from Google summarizer is low hanging fruit.