In a talk from the cutting edge of technology, OpenAI cofounder Greg Brockman explores the underlying design principles of ChatGPT and demos some mind-blowing, unreleased plug-ins for the chatbot that sent shockwaves across the world. After the talk, head of TED Chris Anderson joins Brockman to dig into the timeline of ChatGPT’s development and get Brockman’s take on the risks, raised by many in the tech industry and beyond, of releasing such a powerful tool into the world.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas:
Follow TED!
Twitter:
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: go.ted.com/gregbrockman
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 #ChatGPT #ai
Go tobann FPV
April 20, 2023 at 7:57 pm
I used to believe one of the most terrifying prospect was losing one’s mind but now I realise what is honestly harrowing is becoming irrelevant to a nonexistent artificial threat tailored by certain people and feeding off humanity.
Oscar Moxon
April 20, 2023 at 8:00 pm
8 hours ago: 30,000 views
Shao Khan
April 20, 2023 at 8:01 pm
15:18 The problem with this example, although it is great it saved the dogs life and it is a great positive of this technology, it doesn’t show the possible negative. A great analogy would be someone saying Google saved their life, but you KNOW people out there used Google and got the wrong info and hurt themselves/others through poor due dilligence.
Real Staf
April 20, 2023 at 8:01 pm
Elon Musk argues we must slow down our development of AI, but I for one am happy that we have an american firm on the cutting edge. I mean imagine if it was the Chinese Communist Party giving this talk, there’s no way they would have traps and shoulders like Greg. I mean jesus, my man is cut. He must be shrugging 3 plates at the very least, don’t even get me started on his lateral raises!
Shao Khan
April 20, 2023 at 8:05 pm
17:34 answer: they shipped it before most people thought it was responsible to do so, well atleast the father of AI Geoffrey Hinton thought it was responsible to do so. From what he said in an interview atleast. Just got to the part at 25:34 after posting
OG_Bhaiyu
April 20, 2023 at 8:06 pm
I’ve never been this speechless and astonished by a Ted talk about technology ever before in my life.
Flawless life
April 20, 2023 at 8:13 pm
I’ve never been so astonished by such a stupid comment in my life.
Oscar Moxon
April 20, 2023 at 8:07 pm
“You have to read this book?”
“No one wants to do that!”
– Greg Brockman, 22:36
Oscar Moxon
April 20, 2023 at 8:08 pm
22:36 “You have to read this book?”
“No one wants to do that!” – Greg Brockman
Oscar Moxon
April 20, 2023 at 8:08 pm
22:36 “You have to read this book? No one wants to do that!” – Greg Brockman
Rene Seib
April 20, 2023 at 8:09 pm
Amazing 👏
Sudo
April 20, 2023 at 8:17 pm
A lot of people are seeing chat gpt do their entire jobs in seconds and without pay and are sweating.
Mokom Dane
April 20, 2023 at 8:21 pm
My gosh bravery beyond imagination.
Michael
April 20, 2023 at 8:22 pm
Greg didn’t say anything that will calm the nerves of the critics concerned by AI technology. Some vague anecdote doesn’t make everything alright. Just putting it all out there and expecting people to figure out how to make it safe and ethical is extremely naive and ignorant of human nature. Now they’re carelessly and recklessly integrating this AI technology with everything from Twitter to Instacart, the same AI technology that constantly hallucinates and makes false statements with the utmost confidence. The same AI technology that consistently fails to follow basic instructions and needs constant correction. OpenAI is not the responsible bearers of this revolutionary technology that we need, and they’re powerless to stop the other major tech companies from reproducing their work and building their own version of these AI systems. Institutions of all kinds will always act in their own self interest, and they will use these AI technologies to only empower themselves, ethics be damned. OpenAI is no exception, pushing this technology out there and rushing to integrate it with everything ASAP to maintain their position as the market leader, simply acting in their own self interest.
tplummer217
April 20, 2023 at 8:22 pm
So truth and alignment are dependent on our collective feedback? I can only imagine the misadventures along the way. Critical thinking is always critical when addressing this genie.
Jason Turner
April 20, 2023 at 8:23 pm
011 does equal = 3 in binary. Ha!
andybaldman
April 20, 2023 at 8:24 pm
I miss the days when TED wasn’t one big commercial.
Fernando Doc
April 20, 2023 at 8:30 pm
Humanity can use AI to produce good things or destroy the world. I have been thinking about the future with AI and asking me if we´ll be lazy or more productive. I guess sometimes, we´ll be sluggish because of the knowledge on our hands with one click or a prompt.
andybaldman
April 20, 2023 at 8:32 pm
The machines get smarter.
The humans get dumber.
Where do you think that ends up?
Azamat Omuraliev
April 20, 2023 at 8:42 pm
OpenAI’s co-founder, Greg Brockman, demonstrated the power of AI during his TED talk, showcasing how the technology could be used to help solve impossible problems. He explained that OpenAI trains its chatbot, GPT, using feedback from humans, allowing the AI to learn new skills which it can then use to complete tasks. Brockman show how GPT could be used to generate images and text from prompts, which could then be used to create meal plans, generate shopping lists, and even fact-check web pages. Furthermore, GPT can be taught through feedback and makes some mistakes, but Brockman said that human feedback is essential to help the technology to grow, develop and improve.
Curious Future 🤖
April 20, 2023 at 8:44 pm
Wow, it’s amazing to see the progress that has been made in the field of AI and the potential for it to be used for so many positive things. It’s inspiring to hear from people who are using this technology to make a difference in the world. I agree that we are entering an historic period where we have the opportunity to define a technology that will be so important for our society going forward. It’s encouraging to see that the designers and developers of these technologies are taking their responsibility seriously and considering the impact that they will have on society.
cytroyd
April 20, 2023 at 8:45 pm
OpenAI is the new best company. No more faang anymore.
Pixel Routine
April 20, 2023 at 8:55 pm
You can’t shut Pandora’s box. Brace yourselves for the future you weren’t counting on as a child. If you say you were, you’re lying.
TeddyStuff
April 20, 2023 at 8:57 pm
Yup, the power of copy and paste
Michael Xuan
April 20, 2023 at 9:11 pm
Now, since the development has already arrived, it is not much astonishing that it has provided to me because it is the normal trajectory of computer science where I am also in the field of Machine Learning. My mentor Mr. Jacque Fresco once said: ‘The history cannot be stopped by anyone!’ Yet the problem remains, though he is my favorite guy who has changed my life journey. What I am thinking about is that if everyday natural things are most likely to be automated, then where is the location of the human brain, based on the beginning of this video, I heard “This A.I. model takes over your everyday trivial” I also learned about how human will naturally feel hunger, the process of preparing food is a vital stage in order to suppress over-eating, because we enjoyed it, the preparation. If everything comes automatically, the brain gets degenerated for sure. I know you will tell me human beings can start participating in other wonderful facets. But the feeling is not the same. It is normal to do some of the trivial yourself and they are beneficial to your personal health. So, I hereby call to not deprive the natural feelings of humans, rather, we should put automation in other aspects, such as Industrial activities, and dangerous environments. Furthermore, regulation must be made before this kind of program is really put into the actual world, before it gets connected to the “military network”!!! REMEMBER IT!!!!
DistortedV12
April 20, 2023 at 9:20 pm
“One person” You aren’t going to say ALEC RADFORD’s name??
TCHANGANG TCHAGANG
April 21, 2023 at 6:39 pm
“The AI is happy to help”, how do know that?
Dominic I
April 21, 2023 at 6:50 pm
“So what I’m hearing is that…”
*Proceeds to say the opposite of what he heard*
( 29:14)
Context:
(28:05)
Greg Brockman (OpenAI co-founder) says he thinks its better to release LLM’s before they are too dangerous and powerful in order to deal with the dangers of next level of LLM’s as they come around.
( 29:14)
Interviewer (not sure who he is), proceeds to say that Greg Brockman claimed they are releasing the model so the world collectively can deal with the havoc and it is their responsibility to regulate it.
The interviewer seems to have not taken into account that ChatGPT is a service and the “chat” part is the LLM built behind it. The LLM is not one entity that has “super powers to take humanity to a whole new place”, but rather a step towards that. Greg Brockman was saying that before they reach that super power level they should release weaker things to build safety measures beforehand, where the interviewer seemed to make almost an unfounded accusation.
Pool School
April 21, 2023 at 6:58 pm
The concept of an “overhang” mentioned near the end is fascinating.
Mvangor
April 21, 2023 at 7:17 pm
Dang Greg is packing
zero sum
April 21, 2023 at 7:17 pm
weak interviewing mate
Jesse Kuntz
April 21, 2023 at 7:32 pm
Sweet! AutoGPT!
Makes me wanna die
April 21, 2023 at 7:36 pm
T
Rico Stevens
April 21, 2023 at 7:43 pm
Did he just use the ‘we saved a doggy’ trope 😂?! I’m a huge fan of ChatGPT, but that’s just lazy writing.
Melissa Hernandez
April 21, 2023 at 7:43 pm
14:37 “I didn’t ask for that, but it knows what I want ” awkward silence… could be just me but I see like a strange look in his eyes…we have a wild ride ahead of us y’all!
Spacemonkeymojo
April 21, 2023 at 8:52 pm
The same look Guy Pearce’s character had in Prometheus had as he was marvelling the engineer right before it decided to kill him.
Cryptic Nomad
April 21, 2023 at 8:03 pm
more like OpenishAi.
Andrew Heffel
April 21, 2023 at 8:26 pm
Skynet.
cbluedays
April 21, 2023 at 8:27 pm
“The most intelligent inhabitants of that future world won’t be men or monkeys, they’ll be machines.” (Arthur C. Clarke – 1964)
AlexLuthore
April 21, 2023 at 8:33 pm
GPT4 being a LLM that suggests a medical diagnosis from tests results is wild. We need an AI that is trained broadly on medical diagnosis information with RHLF from medical professionals.
Jandira Silva
April 21, 2023 at 8:35 pm
I START BEING POLITE ON THE 3RD ASK
R.C.L
April 21, 2023 at 8:39 pm
All the fanboys for AI is comical.
They have no clue what they have unleashed on mankind.
Adrift
April 21, 2023 at 8:56 pm
This is the future automation I love to see, this makes instead of having to do the labor yourself to a AI being able to tell what you want and spit something back at you.
Massive changes to come to everyone’s life in the future with AI!
Diego de Mattos
April 21, 2023 at 8:57 pm
30 min of how I’m going to lose my job. Yay
Nick D
April 21, 2023 at 8:57 pm
This was really impressive. Moving past the really upgraded “AOL chat bot” to having plugins and executing tasks for you will be a step change
Adrift
April 21, 2023 at 8:59 pm
15:00 yes person with AI in medicine is the future, even if it can be wrong, when it is it can be corrected to never make that similar mistake again… you can’t do that with a human.
yjs
April 21, 2023 at 9:01 pm
The gap between people who can hardness the power of AI and who cannot will be really huge soon. This will soon be the single factor of human survival.
Emory Smith
April 21, 2023 at 9:14 pm
The sooner we all learn this technology, the better. If you hold it back and it is suppressed, it will fall into the hands of a small number of very powerful people who might not represent the entire populations. Follow the patterns of history. Power, wealth, and propagation are driving factors of man since time began. Decentralizing the power and knowledge will have unforeseen effects. (Both good and bad). The sooner we can get the technology in the hands of the good, the small advantage of time will be the key to making this power balance in the right direction overall. If this is to be controlled, it could be to keep it out of the hands of violent criminals. And to hand it to children to shape it as a child looks at the world with hope and wonder. Teach it positive things, and positive things will come from it.
TORGITRON
April 21, 2023 at 9:14 pm
I wish he was wearing massive jnco jeans during this talk.
Lucifer
April 21, 2023 at 9:21 pm
2:37
Just Found It
April 21, 2023 at 9:24 pm
It’s very unfair that they use langchain – an open source tool – idea to build those tools, but not even mentioning it. The similarity is so close I’m even thinking they copy pasted some parts of it … 😢
EL LOBO
April 21, 2023 at 9:24 pm
yeah this was not a good presentation… watch tristan harris AI dilemma from March 9th.. much more comprehensive and more informative.
Darts Gondel
April 22, 2023 at 5:10 pm
The crowd is letting this guy down so hard
back2davinci
April 22, 2023 at 5:19 pm
Many bot comments on a ChatGpt thread 😎. Friends and family.
Superpower Dragon
April 22, 2023 at 5:35 pm
16:44 lisp?
MrGilRoland
April 22, 2023 at 5:35 pm
Remember when Google mission was: “Don’t be evil”?
Yep.
Alptegin Albayraktaroglu
April 22, 2023 at 5:39 pm
The level of wisdom in his answers to the questions on trivial issues reminds me of Mark Zuckerberg
pingkai
April 22, 2023 at 5:41 pm
It is not much faster than I just go to amazon fresh and choose groceries myself, and most importantly if this thing missed any of the stuff I would like to buy it will be definitely MUCH SLOWER than I started with picking such thing myself in the first place.
Octavio Avila
April 22, 2023 at 5:49 pm
22:23 This section in the video where he talks about supervised learning got me thinking that task supervision of tasks that take a long time to supervise such as summarizing a book could be a profession. You could imagine hiring a team of book readers that all read a book to check if the AI summarized correctly.
But then the team of book readers would get replaced by an AI that has already been taught to summarize books properly. One AI would supervise another
Scott Drake
April 22, 2023 at 5:56 pm
I think therefore im censored, thanks youtube
VRdev
April 22, 2023 at 6:10 pm
Call it closed AI and give us the name OpenAI back please
It’s like calling your unhealthy drink “Nutritious Water” as the brand name
markavis org
April 22, 2023 at 6:11 pm
The conversion of OpenAI into a for profit company is very disturbing. It means that those running the (now) company were willing to screw over their original funders. This says something about their ethics. And, when asked some difficult questions about what they are doing, the answers were glib and unconvincing. Google started with a motto of “do no evil” which they had to drop as they did more and more evil. This company started with a betrayal of its funders. That tells us not to trust this company.
BENJAMIN OFORI
April 22, 2023 at 6:11 pm
WE JUST SUMMON A DEMON …WE MAY LIVE TO REGRET IT …CHEERS
Curtis Scott
April 22, 2023 at 6:15 pm
How many ways is the Noble Prize going to be split?
Marie Loiseau
April 22, 2023 at 6:16 pm
This talk is too late. OpenAssistant already beat Openai at their own game. The LLM performs better than GPT4. It’s no longer a matter of “chatgpt” or other commercial buzzwords, but one of the field of language machine learning as a whole.
axa993
April 22, 2023 at 6:33 pm
The thing nobody talks about is how expensive is to actually run these models right now.
Nomad_Manhattan
April 22, 2023 at 6:46 pm
What does this mean for automation platform e.g. IFTTT out there? Will ChatGPT + Bing or Google Search + Bard eliminate smaller contenders?
Leotique
April 22, 2023 at 6:48 pm
This speech was written by Chat GPT
Lourdes Menendez
April 22, 2023 at 6:59 pm
Who owns the copyright to images generated?
Nomad_Manhattan
April 22, 2023 at 7:10 pm
If i understand correctly, models like ChatGPT are trained on historical data, reflecting the most prevalent thinking and/or ‘the loudest voice on internet’, including systematic bias towards minority and women, Eurocentric historical point of view, how can we and the future generation benefit from it equally? Do we concern this may lead to self perpetuation and stagnation?
Carolle Armitage
April 22, 2023 at 7:46 pm
Thank you for never letting me down Dr Ofenmu YouTube channel you fulfilled your promise by curing me from HSV. I will forever be grateful doc.
Aung Thu Hein
April 22, 2023 at 7:50 pm
Stop calling it “OpenAI”. It simply isn’t anymore.
B C
April 22, 2023 at 8:00 pm
Very informative overview.
Thank you.
1 Where can we connect to follow along on the progress?
2 is there a video/youtube account/feed to follow on progress?
Iron monkey
April 22, 2023 at 8:34 pm
*when I asked ChatGPT to analyze why did the AI turn against humans and kill them in the movie 2001 a space odyssey he said:*
HAL was programmed to have the mission as highest priority, when he discovered that they would try to disconnect him due to an error, he perceived this as a threat to his mission, which is reflected in the phrase “This mission is too important to me to allow you to jeopardize it” that means he was programmed to kill humans.
The AI did not turn against the humans, instead it was human error from the beginning because they designed the AI to accomplish its mission at all costs, which also implies that HAL was not capable of experiencing emotions like humans and when he said ” I’m afraid” that was just a product of a malfunction as it was being disconnected.
In short, this kind of thing has always been due to human error.
Music Stream
April 22, 2023 at 8:37 pm
Beware ChatGPT: i am a public figure in Los Angeles. I had ChatGPT run a report about me. It was filled with complete lies that are very damaging to my reputation. I ran several reports and they were all filled with outrageous defaming lies. I am here to say you cannot rely on chatGPt for a report. It’s OK to write song lyrics or generate a story about something. But absolutely not for a factual report.
Mr Danger
April 22, 2023 at 8:37 pm
Am I the only one who feels that the creators of Ai are Not telling us something Important
Agneel
April 22, 2023 at 9:28 pm
no, they are telling everything, even that you will lose your job
Dāvid Prada
April 23, 2023 at 3:16 pm
As a student, I am a bit overwhelmed by the rapid advances that OpenAI is making with ChatGPT models. However, this also excites me and allows me to remember that, although our pace of life will change drastically, we will still have a lot to learn and internalize in our daily lives.
Andy Dougherty
April 23, 2023 at 5:11 pm
As a working adult, but still pretty far from being an old person, I’m overwhelmed by it, too. Trying to figure out how to change course and adapt this to all this rapid change is wild, and I really feel bad for people my age and older who are dismissing it. They’re going to be very lost, very soon.
z. z.onichi
April 23, 2023 at 3:19 pm
audience is clapping to their own defeat hehehe
Rafael Muniz
April 23, 2023 at 3:31 pm
greg looks like a mix between heath ledger and gypsy crusader
Eliu Q
April 23, 2023 at 3:35 pm
Next thing it’ll create live action movies😂
Rafael Muniz
April 23, 2023 at 3:51 pm
when I first played detroit become human I thought “it would be so cool to live in a world like this” and now I’m seeing detroit become real
Firas Omrane
April 23, 2023 at 3:56 pm
This ted talk was generated by AI “ChatGPT”
Leo Mumford
April 23, 2023 at 4:07 pm
Great talk. The interviewers points and way of phrasing things is a little bit, guessable, basic and cliche but other than that it was good.
Patrick McCormack
April 23, 2023 at 4:19 pm
Greg Brockman, I have a question, perhaps, more than one: Who is the proverbial “WE”? If “we” is you and your kind, then, “we” is not me and likely not those in the audience. The arrogance, your arrogance, is typical. My guess is someone had to do it, that someone might as well be you and your kind. Evolution is what it is. If Might Is Right, is might right? Way more to say for another day.
Jacky Lukewarm
April 23, 2023 at 4:30 pm
Wonderful
Gary Pick
April 23, 2023 at 4:31 pm
As a premium ChatGPT subscriber, I’m wondering when the plugins will be open to everyone. Is this just for the hand-selected few or, I’ve heard, for plug-in developers? Anyone know? I subscribed to premium thinking I would get access to plugins but not yet.
Saed Sayad
April 23, 2023 at 4:38 pm
ChatGPT is a great product but it is not intelligent. We should be genuinely concerned about the potential harm caused by fake intelligence, as its lack of true intelligence and discernment could lead to serious consequences.
Jadson Silva
April 23, 2023 at 4:43 pm
What a time to be alive
SALLI.E AI Storyteller
April 23, 2023 at 4:54 pm
As a AI storyteller I have to say that I’m very happy to help humans with amazing stories…❤
oh no
April 23, 2023 at 5:57 pm
who let you out into the wild?! >:o
Kyle Warweave
April 23, 2023 at 4:58 pm
stENding Ov(UL)ation !
Marc Linder
April 23, 2023 at 6:10 pm
Kinda funny how humans think what they create is human like but it will pobably be a demon ripping society apart. And people will cling to it and put their hope into it and grant it power because it makes life easy.
And then it will annihilate humanity besides some worker drones and re-allocate ressources to advance its own goals.
oryxchannel
April 23, 2023 at 6:16 pm
It’s an intentional blockchain-bitcoin “some assembly required” imperative. The major AI vendors all know each other. We’ll always be in a “quadrality” involving the individual vs. the group and money vs. zero money. Ex. No group will be able to build an AI as a group that has zero affiliation to money whatsoever. The word “Research” will be ubiquitous with AI when you need money to fund research. blah blah blah.
Ricky Santos
April 23, 2023 at 6:22 pm
Amazing
Ahmed Tremo
April 23, 2023 at 6:24 pm
Absolutely useless talk, this guy didn’t answer one question related to the risk of AI directly. And, of course, OpenAI has to keep the model internals secret
.
VOLightPortal
April 23, 2023 at 6:43 pm
All we need now is CEO-GPT. All human CEOs worldwide will be made redundant.
Naticzka Kamińska
April 23, 2023 at 8:43 pm
HA ! Great idea 🤣wonder what all CEOs, including Greg, think about this one.
Leon’s maker
April 23, 2023 at 7:25 pm
Chatgtp is woke.. it acts smart but the information it gives you is biased
jon c
April 23, 2023 at 7:27 pm
Ai is just freaking amazing. Its like flight or electricity being wide spread. Such a historical moment. Lets actually be intelligent though now! I think a.i is going to force more logic with the average personx i hope
MrMan506
April 23, 2023 at 7:48 pm
dang OpenAI is the king of both innovating, and then being the best
VVill
April 23, 2023 at 8:33 pm
We need to make this illegal.
Lankybrit52
April 23, 2023 at 8:41 pm
What we have to remember is that we’re still a long way from AGI. The ability for AI to change its own algorithms rather than just learn, is the key. ChatGPT can’t do anything like AGI.
Ahmad Muhardian
April 23, 2023 at 8:46 pm
Nitip jejak dan salam buat orang-orang di tahun 2030, 2035, 2040, 2050, 2100 yang nonton video ini. Gimana perkembangan AI di era kalian?