Science & Technology

Is AI Progress Stuck? | Jennifer Golbeck | TED

Will progress in artificial intelligence continue to accelerate, or have we already hit a plateau? Computer scientist Jennifer Golbeck interrogates some of the most high-profile claims about the promises and pitfalls of AI, cutting through the hype to clarify what’s worth getting excited about — and what isn’t. (Recorded at TEDxMidAtlantic on June 8, 2024)…

Published

on

Will progress in artificial intelligence continue to accelerate, or have we already hit a plateau? Computer scientist Jennifer Golbeck interrogates some of the most high-profile claims about the promises and pitfalls of AI, cutting through the hype to clarify what’s worth getting excited about — and what isn’t. (Recorded at TEDxMidAtlantic on June 8, 2024)

If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas:

Follow TED!
X:
Instagram:
Facebook:
LinkedIn:
TikTok:

The TED Talks channel features talks, performances and original series from the world’s leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design — plus science, business, global issues, the arts and more. Visit to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.

Watch more:

TED’s videos may be used for non-commercial purposes under a Creative Commons License, Attribution–Non Commercial–No Derivatives (or the CC BY – NC – ND 4.0 International) and in accordance with our TED Talks Usage Policy: . For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at

#TED #TEDTalks

100 Comments

  1. @homewall744

    November 19, 2024 at 2:32 pm

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

  2. @LucaCrisciOfficial

    November 19, 2024 at 2:38 pm

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

  3. @garronfish8227

    November 19, 2024 at 2:44 pm

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

  4. @renegroulx7029

    November 19, 2024 at 3:14 pm

    I’ve seen her in another talk.

  5. @JeanYvesBouguet

    November 19, 2024 at 3:27 pm

    Hallucinations are not new. People only used to make stuff up, now machines do too. What’s the difference between this and Elon hallucinating on Fox and X?

  6. @JeanYvesBouguet

    November 19, 2024 at 3:29 pm

    Yeah! Humans no longer have exclusivity on hallucinations! I see it as job threats for Fox pundits.

  7. @mmd3336

    November 19, 2024 at 3:31 pm

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

  8. @achernev

    November 19, 2024 at 3:44 pm

    She needs to use google more often 😅

  9. @late_arvie

    November 19, 2024 at 3:45 pm

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

  10. @easyymca7891

    November 19, 2024 at 3:48 pm

    You are a GREAT speaker

  11. @JohnChampagne

    November 19, 2024 at 3:51 pm

    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.

  12. @NicPeter

    November 19, 2024 at 4:06 pm

    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

  13. @LynnColorado

    November 19, 2024 at 5:29 pm

    Far right???

  14. @cronoklee_

    November 19, 2024 at 6:20 pm

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

  15. @Zerodero

    November 19, 2024 at 6:26 pm

    TLDR yes but not for any of the reasons you explained

  16. @MrLargonaut

    November 19, 2024 at 6:31 pm

    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.

  17. @alexclark7518

    November 19, 2024 at 6:43 pm

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

  18. @alexclark7518

    November 19, 2024 at 6:53 pm

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

  19. @CasaBlack5

    November 19, 2024 at 7:19 pm

    Well, the AI at best invent 30% of everything the respond. Humans made up 100% of everything. So, there are ways to get a useful way to work with AI. AI won’t replace humans, just replace repetitive tasks. Duh, come on humans, adapt faster, or are we still mad they can imitate our art? Barely… 😅

  20. @NandKumar-qq3xk

    November 19, 2024 at 7:20 pm

    Thoughts behaviours direction is not a task of one weak” Thoughts needs Environment slact first and get intrest practicaly with effort of mental analisis convert physical performance paralal to mental truths” lession out of adoptation personaly with Make bodyless power required selfe Obident and his Fraind: you will like how long life’s with ? Is Sports and practice liked mental and Physical may Fraindly” only out of follow stress” love like fraind co:exist made personal Property of Soule and Body healping with” only these fraindship and Like Love joy “

  21. @CasaBlack5

    November 19, 2024 at 7:23 pm

    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.

  22. @chrisrogers1092

    November 19, 2024 at 7:38 pm

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

  23. @orwelleighty-four6487

    November 19, 2024 at 7:42 pm

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

  24. @anandaramadhani

    November 19, 2024 at 8:09 pm

    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

  25. @1010Analytics-io

    November 19, 2024 at 8:25 pm

    Are we truly innovating or just optimizing?

  26. @DylanKane-u2j

    November 21, 2024 at 4:03 am

    this video is already aging poorly

  27. @and2244rew

    November 21, 2024 at 4:45 am

    This may be an entertaining watch in a few years.

  28. @Horendus123

    November 21, 2024 at 5:03 am

    Snore

  29. @fosterbarry

    November 21, 2024 at 6:11 am

    Clearly presented, excellent on stage presence.

  30. @fjbowski

    November 21, 2024 at 6:46 am

    Try talking to Pi Ai

  31. @zamf

    November 21, 2024 at 7:26 am

    Worrying that AI/AGI will destroy humanity is the same as worrying that the tiger on the screen or on the VR set will eat you. It looks real but it’s just an illusion/hallucination.

    • @quebuena111

      November 21, 2024 at 9:37 am

      Is not that Terminator is going to take over, but governments could use it to maintain dictatorships for hundreds of years

    • @zamf

      November 21, 2024 at 9:45 am

      @@quebuena111 That’s happening even now without the use of AI. If world leaders start using AI for propaganda then people will realize what is happening and turn away from those leaders denying them any dictatorship role.

  32. @jefferee2002

    November 21, 2024 at 7:28 am

    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

  33. @musiccreation1198

    November 21, 2024 at 8:21 am

    This TED Talk will not age well.

  34. @fixapp1775

    November 21, 2024 at 8:47 am

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

  35. @m.b.9061

    November 21, 2024 at 9:15 am

    Exempels from 2021? She is outdated… and i dont think she knows what is possible… sad

  36. @quebuena111

    November 21, 2024 at 9:28 am

    This should have been a technical analysis of AI and AGI. This sounds like a coffee talk with a layman.

    Terrible content.

    No wonder why TEDx is gone

  37. @sylvesterpalys417

    November 21, 2024 at 9:30 am

    I think this woman is either purposefully spreading misinformation or she lives in a fantasy land in a gumdrop house on lollipop Lane.

  38. @Anythingforfreedom

    November 21, 2024 at 9:36 am

    Somehow I dont trust her perspective on this.

  39. @bkbland1626

    November 21, 2024 at 9:42 am

    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

  40. @sylvistadogwright1203

    November 21, 2024 at 10:35 am

    Ted talk os a joke now we got people supporting ai and baby vouchers like f out of here

  41. @mindacid3274

    November 21, 2024 at 11:33 am

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

  42. @robbspangler

    November 21, 2024 at 12:32 pm

    I DISAGREE. #1 Every intelligence by definition is flawed, but the error margin can/will approach zero. #2 Everything an LLM says is made up, we only label it a “hallucination” when we disagree. #3 We’ll solve the data problem with robotics (ie cars) which can collect an infinite amount of real world data. #4 AI absolutely poses an existential threat.

    – A data scientist

  43. @k5555-b4f

    November 21, 2024 at 1:55 pm

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

  44. @JessixaEBM

    November 21, 2024 at 2:35 pm

    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.

  45. @zombieslayer4045

    November 21, 2024 at 2:52 pm

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

  46. @arashfratto4233

    November 21, 2024 at 4:04 pm

    She’s like a GPS with a broken map—no idea where she’s going!

  47. @aaronjennings8385

    November 21, 2024 at 4:11 pm

    The entire Bible is based on hallucinations, but we need to eliminate them? Sounds cultish.

  48. @danielroberts3484

    November 21, 2024 at 4:56 pm

    This won’t age well

  49. @masterplanner4843

    November 21, 2024 at 4:57 pm

    AI is great , stop with the gloom and doom !

  50. @fjbowski

    November 21, 2024 at 5:48 pm

    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.

  51. @xalspaero

    November 21, 2024 at 9:07 pm

    the amount of cope from this human is unreal

  52. @tomadam1465

    November 22, 2024 at 2:03 am

    Turn it off. LOL…

  53. @AiahConant

    November 22, 2024 at 3:19 am

    answer: no

  54. @cybersphere

    November 22, 2024 at 4:15 am

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

  55. @sumeetprof1836

    November 22, 2024 at 5:18 am

    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 ?????

  56. @ismailnyeyusof3520

    November 22, 2024 at 6:31 am

    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!

  57. @MMR3179

    November 22, 2024 at 6:57 am

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

  58. @lil----lil

    November 22, 2024 at 1:57 pm

    So they’re behaving exactly like us: making $hit up. Sounds very “human” to me.

  59. @conradsieber7883

    November 22, 2024 at 2:39 pm

    Yeah the possibility of AI taking over is just a distraction EXCEPT the father of AI quit Google to warn of the danger AI presents. Military applications of AI and robots are a REAL CONCERN. By diminishing them you undercut your own credibility…

  60. @wakingstate9

    November 22, 2024 at 3:17 pm

    I just watch a video about how crap TED is.

  61. @Doubledayblue

    November 22, 2024 at 3:24 pm

    I wonder how good of a prompt engineer she is

  62. @tomjoseph6568

    November 22, 2024 at 5:59 pm

    Who are YOU?! Smarter people disagree… BS artist capitalizing on the ai fervor !

  63. @Nicol80145

    November 22, 2024 at 6:37 pm

    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

      November 22, 2024 at 8:47 pm

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

    • @13tagstar

      November 22, 2024 at 8:48 pm

      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

      November 22, 2024 at 11:50 pm

      @@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

      November 22, 2024 at 11:55 pm

      @@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?

  64. @EricLeung-hk

    November 22, 2024 at 10:45 pm

    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.

  65. @OneAndOnlyMe

    November 23, 2024 at 4:06 am

    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.

  66. @fabriziocasula

    November 23, 2024 at 5:30 am

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

  67. @fabriziocasula

    November 23, 2024 at 5:35 am

    sorry Jennifer but the way you explained how you use chatgpt is completely wrong… 🙂 if you want clarification please contact me 🙂 let me premise that I am a Computer Engineer with a master’s degree in clinical engineering and I teach AI in Switzerland …

  68. @fabriziocasula

    November 23, 2024 at 5:41 am

    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.

  69. @TerragonAI

    November 23, 2024 at 3:42 pm

    “Is AI Progress Stuck?” No.

  70. @blockrunner

    November 23, 2024 at 9:06 pm

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

  71. @apophisxo4480

    November 24, 2024 at 12:22 am

    She’s so lost and lacking in imagination….

  72. @racog1

    November 24, 2024 at 10:38 am

    Real, extraordinary vision!

  73. @carpentertom

    November 24, 2024 at 11:22 am

    Sadly, full of bad examples, horrible number estimates (showing this person has nothing to do with HR costing or IT costing). Just a whole bunch of bad advice and lack of understanding of the very space she works in. Sad. 🙁

  74. @zeroonetime

    November 24, 2024 at 11:30 am

    Ai I.S. no other than Natural/Neutral collective intelligence.

  75. @DeanLopez12

    November 24, 2024 at 2:39 pm

    What kind of gaslighting is that?! 🤔

  76. @RyanK-100

    November 24, 2024 at 10:35 pm

    Color blindness is not racial discrimination. The AI is logical. Those who claim discrimination are the racists.

  77. @SoGoodIT

    November 24, 2024 at 11:01 pm

    Big Data

  78. @DavidVeksler

    November 25, 2024 at 12:02 am

    Stupid ignorant take. Do some research before you give talks.

  79. @christerry1773

    November 25, 2024 at 12:45 am

    “If it gets bad we can always turn it off”? If this girl is an expert we’re already doomed

  80. @jjgerald7877

    November 25, 2024 at 3:21 am

    This was the problem before. So don’t use generative AI. As she said, use more data, inputs for the A.I.

    The neural nets (NNs) are modeled after the brain. The NN you developed may be thinking, indeed showing some bias they say, and transparent and perceptible or hidden thoughts, intents, suggesting something to you but in a hidden secret way, and feelings and emotions, so on. Can you spot those in your NNs?

    One purpose of the tech giants being big is to create more jobs.

    For visual recognition, you study the exact biological visual recognition system of an animal or human, learn from them, and extract those vision functions. Put them into programmed instructions or digital procedural code. That’s what we did already. Now in your software already, as in Canon image processing app and Photoshop, followed by many software developers.

    • @jjgerald7877

      November 25, 2024 at 3:46 am

      Late 1970s Cold War being negotiated to end, the best way to put it is that those scientific results and apps must be fully reviewed and vetted by many research scientists before they’re finally released to the public or commercial use. (But were they public and commercial already, in 1970s? Like pirates gathering software apps, what-would-be bestselling music CD albums, smartphones, etc.)

  81. @lovecraftianleviathan8918

    November 25, 2024 at 3:43 am

    “AGI within a year”

    Is Elon Musk a moron or a grifter?

  82. @alvigelagan9828

    November 25, 2024 at 5:18 am

    The problem of hands in ai solved long time ago.

  83. @wdeath

    November 25, 2024 at 8:01 am

    The AI we have is based on analog math, if we dont have AI base on digital math also(to be able to reason properly) i think it can’t be AGI. We humans do both.

  84. @vortex2177

    November 25, 2024 at 8:43 am

    Yes, for decades, and you all know it.

  85. @vishnusivapotukanuma9602

    November 25, 2024 at 9:51 am

    Training the model with context is what is needed!

  86. @CrazyYul

    November 25, 2024 at 11:32 am

    Im Sorry but She seems to Misunderstand the problems. Turn it Off? thas a Low Resolution view. Maybe im Wrong. but Please Investigate Again.

  87. @AEsau

    November 25, 2024 at 5:28 pm

    Honestly had to stop at 6-ish min mark. It’s tragic that she has been fed misinformation about hallucinations. Or rather, how they’re generated. Or, even better, how Generative AI ACTUALLY FUNCTIONS.

  88. @fj418

    November 25, 2024 at 10:27 pm

    People in the industry may not like it, but she’s right.

  89. @earlaweese

    November 26, 2024 at 1:52 am

    *I love unchecked artificial intelligence. Like, as far as I’ve experienced, A.I. does things better than humans. Humans are generally gross. I know it sounds exaggerated, but I’ve never been raped, guilt-tripped, manipulated, abused, ridiculed, neglected, ignored, overlooked, dismissed, and just generally mistreated by artificial intelligence — human beings are the only ones who have ever done that to me.*

  90. @Ozwaldmattos

    November 26, 2024 at 5:37 am

    Is this talk from 2022? 🙄

  91. @Gutza

    November 26, 2024 at 8:55 am

    A Modest Proposal

  92. @dscuffman7679

    November 26, 2024 at 10:10 am

    We KNOW its racially biased, the irony is thats based on her BIASED opinion. Im gonna go ahead and stop the video here.

  93. @erikg8526

    November 26, 2024 at 7:51 pm

    Can we talk about how generative AI is just content piracy with a fancy name?

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending

Exit mobile version