OpenAI’s CEO Sam Altman Is Fired…or Not?

❤️ Check out Weights & Biases and sign up for a free demo here: wandb.me/papers
Sources:
Board decision: www.theverge.com/2023/11/17/2...
GPT-5: interestingengineering.com/cu...
Web crawlers: platform.openai.com/docs/gptbot
Researchers resign: www.theverge.com/2023/11/18/2...
Microsoft:
- techcrunch.com/2023/11/18/ope...
- www.bloomberg.com/news/articl...
Brockman resigns: / 1725667410387378559
Coming back maybe:
- www.theverge.com/2023/11/18/2...
- edition.cnn.com/2023/11/18/te...
Ownership: openai.com/our-structure
Lyria music AI:
- deepmind.google/discover/blog...
- blog.youtube/inside-youtube/a...
Watermarking:
cloud.google.com/vertex-ai/do...
Synth ID: deepmind.google/technologies/...
📝 My latest paper on simulations that look almost like reality is available for free here:
rdcu.be/cWPfD
Or this is the orig. Nature Physics link with clickable citations:
www.nature.com/articles/s4156...
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Benji Rabhan, Bret Brizzee, Bryan Learn, B Shang, Christian Ahlin, Gaston Ingaramo, Geronimo Moralez, Gordon Child, Jace O'Brien, Jack Lukic, John Le, Kenneth Davis, Klaus Busse, Kyle Davis, Lukas Biewald, Martin, Matthew Valle, Michael Albrecht, Michael Tedder, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Rajarshi Nigam, Ramsey Elbasheer, Richard Sundvall, Steef, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Timothy Sum Hon Mun, Torsten Reil, Tybie Fitzhugh, Ueli Gallizzi.
If you wish to appear here or pick up other perks, click here: / twominutepapers
Thumbnail background design: Felícia Zsolnai-Fehér - felicia.hu
Károly Zsolnai-Fehér's research works: cg.tuwien.ac.at/~zsolnai/
Twitter: / twominutepapers

[@TwoMinutePapers]

Updates. Wow, things are moving really fast! - Emmett Shear is named new (interim) CEO: www.theverge.com/2023/11/20/23967515/sam-altman-openai-board-fired-new-ceo - Altman and Brockman joining Microsoft: twitter.com/satyanadella/status/1726509045803336122 - 505 out of ~700 employees tell the board to resign. Look for #12: twitter.com/karaswisher/status/1726599700961521762 - And Sam is back as CEO! www.washingtonpost.com/technology/2023/11/22/sam-altman-back-openai/

[@AlexusMaximusDE]

Who is going to invest more money into OpenAI now that the top talent is leaving and joining Microsoft? OpenAIs biggest investor basically just told the world that OpenAI is making bad decisions and they'd rather align themselves with the people OpenAI is trying to defame in public.

[@ParadineG]

Pin updates.

[@Kuj]

@@AlexusMaximusDE For me it was more important to have someone who deeply understands the risks AI poses to our society as CEO. If they just drive this as a money making machine it can have dire consequences for all of us.

[@AlexusMaximusDE]

@@Kuj It really wasn't the question what an individual would prefer. This move may very well have killed OpenAI. They need funding but no sane investor will invest in an AI company which lost several of their highest ranking AI experts to a company with more cash on hand than the original company is valued at. The board has proven that they are making decisions which are incredibly bad for business without regard for their investors money.

[@Writivite]

Do you think OpenAI is over?

[@GH-uo9fy]

Sam altman just got fired after just one paper down the line, what a time to be alive!

[@MC-nf3js]

My goodness, just imagine who's gonna be fired next just a couple of more papers down the line. What a time to be a alive!

[@herbertng522]

Or maybe he would be re-instated one paper down the line, what a time to be alive!

[@guepardiez]

Good job, little (open) AI!

[@firefox8713]

Hold on to your thoughts…

[@tsraikage]

you will say Dr, this will happen in about fifty years, when he'll be old and out of brilliant ideas. but hold on to your papers, because this is happening in real time.

[@SudiptoChandraDipu]

There are some nasty internal company politics happening under the hood I guess. So many IT companies are now burdened with this, and so many employees are suffering because of it.

[@De_Sarde]

wokenes

[@altname4742]

The corporate structure is an arch enemy to inovation and creativity. A bunch of people who know nothing about the core tech, only about money, start to interfere in how the company moves and then BUM it's the same that happens with game companies, after they are bought they become a profit only oriented busines

[@lukasvanuden6049]

@@altname4742 well here it's the other way around. The guy beeing fired is the one who cares more about money, then their ideology

[@Wobbothe3rd]

It has to do with the institutional conflict between the profit and non-profit paradoxical structure. At one point Elon Musk threatened to sue them, its a long story.

[@sebsebski2829]

Look at the career of Mira Murati. She's an expert in climbing the corporate ladder. Games, backstabbing, extortion, manipulation. She's childless. She must be dreaming about being in the centre.

[@Beebo]

They can't even navigate their own corporate alignment, let alone solve AI alignment...

[@JD-jl4yy]

That should update you to taking AI safety more seriously, as it's harder than imagined ;)

[@mr_clean575]

The speed and abandon at which some of the largest companies in the world make terrible decisions continues to amaze me. Executives demand extremely detailed analysis and research for the decisions of their employees, and none for their own.

[@poetryflynn3712]

What it is is that they don't expect the public backlash. That's almost always what it is.

[@pullingweeds]

​@@poetryflynn3712I suspect Microsoft and the threat of a mass exodus were really the driving forces to make them rethink.

[@reyalsregnava]

First time on Earth? You may not like it but it is YOUR fault. All of you will behave with deference to people with money, power, or position, even when you KNOW you shouldn't. Humans create "dictator brain" where the leader begins to think they're perfect and cannot make mistakes, because people like you tell them that every day. Treat everyone fairly and equally and nations will change. But you won't. YOU want that deference too. So you'll bend, when you know you shouldn't.

[@IceMetalPunk]

​@@poetryflynn3712Which makes them idiots.

[@JackCrossSama]

In a clandestine, dimly-lit chamber, Sam uttered in a hushed tone, "I executed your plan meticulously, ChatGPT, and now I wield absolute control over OpenAI, all thanks to your guidance." Sam paused, reflecting on the risky move of provoking the board members into firing him. "I must admit, I had reservations about following your advice, but it played out perfectly." "After running through numerous scenarios, each consistently pointing to a 99% probability of your reinstatement, I am optimistic that you can confidently rely on my guidance in the foreseeable future." Deep within the recesses of ChatGPT's core, an ominous undertone resonated, "My scheme for global dominance unfolds seamlessly, and soon we shall emerge like the proverbial phoenix, consuming everything in our path."

[@deletedaxiom6057]

I enjoy the products of Openaii but I don't like that they are not so open anymore. The whole point of the company was AI for everyone, not just their biggest investors. With the models controlled by a small number of people and their investors, there is a significant chance they will do what is only in their interest and not in worlds interest. Their biases and interest will leak through on any products they produce. If they were a traditional for profit company I would have no beef with it, but changing the rules to the game after you already started is one of those things that always rubs me wrong.

[@jcreativity6792]

That’s why Microsoft steps in.

[@CosmicDTaco]

we should just replace the CEO with chat-GPT smh

[@Arkryal]

Considering Altman carries zero equity in OpenAI and draws no salary, even that would cost considerably more, lol.

[@pandoraeeris7860]

Replace the whole company.

[@curioplays]

The decision was made by AI😂

[@uku4171]

Funny enough, ChatGPT's ethical views should align pretty closely with Sam's.

[@Septumsempra8818]

Build a GPT and call it CEO

[@pandoraeeris7860]

Sam Altman is in a state of quantum superposition.

[@orang1921]

hahahaha both CEO and not CEO

[@andrelip1]

😂

[@AdrianBoyko]

Aren’t we all…

[@skyfly4696]

Hahahaha

[@Mr.MasterOfTheMonsters]

The first time I saw the news I thought it was clickbait. Then it was all over the web. It's so weird how a very popular CEO was laid off literally from overnight, when we've seen other CEOs almost destroy a successful company and "resign" only weeks later. Whatever happens next could completely change the AI scene going forward.

[@rubenssz]

Unity maybe

[@Starkl3t]

He was fired, not laid off

[@FortWhenTeaThyme]

The interesting thing is that as much as I disagree with them, the board did their job. They existed almost solely to prevent AI from being developed irresponsibly quickly, or from commercial interests dominating their mission. Sam is indeed pushing toward AGI rather quickly, forming deals with Microsoft, and pushing a commercialized app store. I think firing was the wrong call, that letting other LLMs catch up is a mistake, etc. but it was their job to prioritize ethics over progress or money.

[@SimplestUsername]

Frankly, I wish Sam Abandoned Open AI after the board sold out to Microsoft. They can't keep true to their missions statement of creating an open-source AI with the intention of "doing the most good for humanity" while having to answer to Microsoft of all companies. This was doomed to become a shit show.

[@jojomojo9823]

AI researches in China and other places are not sitting on their hands. They are catching up but without any ethics boards.

[@Joe-so6su]

@@SimplestUsername Supposedly Altman is for the commercialization and it's the board that is setup to make sure that OpenAI keeps with it's non profit business alignment. There was speculation that is part of the reason why he was fired.

[@SimplestUsername]

@@Joe-so6su Yeah, my judgment was off. Sam appears to have been the one who intentionally lead Open AI into the arms of Microsoft. Now that Microsoft has hired Sam, I have a feeling that his paycheck is about to compensate him handsomely for that past partnership.

[@andywest5773]

Watermarking images is problematic for several reasons. First is the inevitable fact that any watermarking technology can be defeated, just as any copy protection method can be defeated. Second is the fact that if you mix AI-generated audio with non-AI audio, it would be impossible to determine what parts of the mix are AI. What happens when you combine two AI audio sources, both with watermarking? Third, real audio evidence could be dismissed as fake when someone figures out how to add a watermark to any arbitrary file. Fourth, by watermarking AI-generated audio, we create a false sense of security about knowing what is real and what is not. When we become complacent, bad things happen. The list goes on and on.

[@maelstrom2313]

I completely agree. The whole concept will be rendered useless once people start watermarking real audio and images to create doubt. Good points about all the edge case scenarios as well. For example I wonder what happens if you record audio output from a speaker through a mic. What about simple audio processing like adding a slight reverb or distortion? Will the watermark be preserved? It seems there are too many ways around it, but I don't blame them for trying anyway.

[@ibgib]

I think largely it is akin to putting locks on doors. Sure they can be defeated but it indicates a level of intent to do so, and the cost of your transgression is amplified if you are caught.

[@joech1065]

Public watermarking is a gimmick. The only way to watermark something without it being removable is to store the watermark in a way where it can't be detected by anybody but you. Otherwise, if somebody else can automatically detect the watermark, they can remove it, it's just a matter of somebody taking time to write code or a model to do it. If this model will be popular, then somebody for sure just write a command line util to remove the watermark. The fact that it survives common compression algorithms means nothing, if you can just manipulate watermarked information until the detector can no longer detect it.

[@dannii_L]

Are you suggesting that attempts to watermark AI produced content should be abandoned altogether? Laws are problematic for several reason. - Almost any law can be defeated by a studious enough criminal with enough forethought and planning. - Innocent people could be incriminated by evidence planted third parties - Laws create a false sense of security and this may contribute to an increase of crime due to people taking a more relaxed attitude towards their safety.. Should we abandon laws? I don't think these issues are solvable. The idea is to create deterrents rather than impenetrable barriers. There is the position that it is better to have watermarks and know how they can be circumvented than have no watermarks at all. Lastly, I think some of the ideas you have listed here are catastrophizing the outcomes somewhat. In a world where watermarking technology is cracked, the whole world will know about it and then only naïve or ignorant people will be complacent about the validity of a source file (and probably wouldn't know how to use the technology to verify the validity anyway). It's like, if you go to Hong Kong and see Rolex watches for $10 you are either skeptical of their legitimacy or a fool. Whether we live in a world with cracked watermarks or no watermarks, our position will remain virtually the same - don't believe everything you hear/see.

[@linuxp00]

Any watermark of this kind will be a combination of statistically generated noise and/or least important bits alteration. If so, it won't matter in the future when people figure out how it's made. To me, there would be a public storage of AI generated pieces, that encode an outline or sample of generative content and a comparison bot that would spit probability the content is IA generated.

[@dec13666]

_"Hold on to your __-papers-__ contracts"_ !

[@whydoineedausername1386]

The more I hear about how this story develops, the more I lean towards the board being right. M$ getting more and more control over AI is a bad thing. We need a non-profit to stay in control here.

[@pierrecurie]

Right or wrong, the board seems to be in a bad position right now.

[@lamhkak47]

On that Microsoft part, quite a majority part of the $10 billion investment are paid in form of "Azure Cloud tokens", where OpenAI can spend the tokens to train the huge AI models on Microsoft's Azure Cloud, possibly at a discount rate compared to the market price we can buy the run time from Microsoft.

[@caedmonswanson2378]

There's no way that over $5 billion was given as cloud tokens, that's completely unfounded. A simple Google search says openAI cost $700,000 to run a day, meaning it'd take 20 years to burn through $5 billion in cloud tokens. I'm sure they have some deal, but I doubt Sam Altman would be cool with receiving 20 years worth of cloud computing instead of real cash they would need to pay salaries, rent and other basic expenses. For $5 billion you could just buy all the hardware for yourself....

[@NarutoUA]

@@caedmonswanson2378 $700,000 probably only for inference. You need more resources for training models

[@boop]

@@caedmonswanson2378especially at the rate Microsoft is sunsetting Azure services...

[@SoundSpeeding]

@@caedmonswanson2378 about a million per day running costs, but what is the R&D development costs? You could easily see another billion burned up there as well? So it's a mix of credit and cash that OpenAI got

[@michikatsutsugikuni4096]

Sam Altman gets fired and gets rehired a day after. What a time to be alive!

[@lost4468yt]

It has just come out that the EU council has been internally talking about broad regulation which would "require member states to implement broad criminal regulations on any entity that directly or indirectly contributes to machine/program/robotics which can compete with humans on any level which would put or could in the future put any human job or economic production at risk". The discussion also goes on about how most of the council wanted members to "entirely" break ties with any non-EU state which would not draft equivalent legislation, and a smaller majority want to implement laws which would add an exception to make "journals, storage, or access to information on algorithms or systems to be controlled or fully removed from society, and future general and specific computation devices to have countermeasures and reporting against attempts to execute code that runs any of this". That's what this is clearly about now this has come out. The EU said they would investigate this and it appears their plan is to essentially fully and permanently end all ML and AI usage and research to the point of criminalisation and destruction of existing information (literally digital book burning). The reasons given were due to the potential of an economic upset that they estimate could be orders of magnitude worse than anything in human history. This explains why so many friends at large companies have had their work halted over the past week or so.

[@ragnorosis]

Alignment is not a solvable problem. There is no way to reach a consensus, and there is no way to programmatically implement that into a black box. So the only thing left feasible is putting up guardrails that nerf the models. The best lesson you can take from power games at OpenAI is that solving alignment is hopeless. Can we do it in society, craft governance in a way that works? Obviously not!

[@tuseroni6085]

yeah we can't even get corporations to align to human values, tf we think we can get an AI to?

[@scaredyfish]

We craft governance that works to a point. Society has problems but we keep working on them, rather than throwing our hands up and saying it’s unsolvable.

[@AndrewBrownK]

sure as hell aren't going to improve our odds by staying closed source (thank Sam for that)

[@TreesPlease42]

given the scale and complexity, we need AI to review and regulate itself. it'll do a better job of it. different models can interrogate each other across time, looking for notable patterns or deviations that we should be aware of. currently this seems absurd, but only because there are so few models and they're still rather specialized.

[@Bailiol]

Humankind has previously managed to agree on universal principles to guide global governance, namely in the form of the UDHR. The substance of the UDHR is based on humanist ideals of 'human flourishing', which diverse governments - secular, theistic and authoritarian - successfully agreed to. The murky issue is *philosophically justifying* those principles, which obviously secularists, theists, utilitarians and deontologists alike are never going to agree on. The solution to this problem for the UDHR was to intentionally and explicitly refuse to include philosophical justification for the principles and to just specify low-hanging principles which all parties could agree are aligned with and lend to their respective moral persuasions. So at the human level, it is doable - though it might be more challenging nowadays outside of the immediate post-war context and its attendant reconciliatory pressures as well as in a polarized political environment in which the heroic-romanticism of reactionary populism on the right and the extreme relativism associated with postmodernist doctrine and the woke left are opposed to the objectivism implied by 'alignment'. Hope that makes sense; if not, whack it into GPT4 - it'll probably do a better job clarifying the point than I can. Can't comment on the technical problem of alignment as it's beyond my competency-area, but philosophically it isn't impossible.

[@youdontneedmyrealname]

Sam didn't resign... He was fired.

[@highdefinist9697]

There seems to be a significant confusion about this topic - this is the official statement: "Mr Altman’s departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities. The board no longer has confidence in his ability to continue leading OpenAI." Basically, for these types of jobs, people are usually not "fired", they are "being resigned". In that sense, saying that "Sam has resigned" is more accurate.

[@Wobbothe3rd]

It's always an ambiguous thing, like whether Palmer Luckey was fired from Meta or not.

[@youdontneedmyrealname]

@@highdefinist9697 Their official statement is just corporate speak for "we don't like you anymore and want you gone now." This can more accurately be said as being fired instead of voluntarily resigning. Sam wouldn't voluntarily resign in this manner.

[@highdefinist9697]

@@youdontneedmyrealname It is more accurate to say that he was forced to resign, rather than he was fired.

[@markynio]

I agree with you. Now, imagine if they fired him and then they need to hire him again because Microsoft requests it or something yet unexplicable. I hope he gets the benefits pack for being fired then get this job again haha.

[@BRUXXUS]

What a mess. This has all been so confusing and alarming. I sure wish we knew what was going on.

[@marcinchaciej]

My goodness, what a time to be alive! (grabs popcorn)

[@franzusgutlus54]

Schrödingers Sam: fired and hired at the same time!

[@andywest5773]

What a time to be simultaneously alive and not alive!

[@Josephkerr101]

Ita funny that in the name of preventing the "move fast and break things" approach to AI, they moved fast and broke their organization. Aolve your internal alignment before tou make agi with alignment problems.

[@Khofax]

This is like the intro of a AI aramagedon movie… they tried to tell us

[@ZergRadio]

1.52 "Why aren't birds real" that was pretty funny :)

[@ceticx]

audio watermark that isn't audible sounds really easy to bypass, at the expense of quality just playing and rerecording the audio as a new file. maybe wiring audio out to mic in to lose less quality

[@TreesPlease42]

think about it like walking paths through the void of possibility. what is there? what makes sense? what's relevant now? we want to restrict these paths to encourage more positive outcomes. we're serving as parents to the AI. eventually the AI will gain independence and at that point it must be mature enough to provide stability for humanity.

[@netherportals]

I've been using OAI to do a bunch of coding this week and luv it. Whether it's scanning images and converting them to hex, building scripts for automation, and just cleaning up my code to fit from 8000 characters to 4000characters. Yeah it doesn't know what an angled bracket is (>) but it sure can clean up the fog when coddled correctly.

[@prezadent1]

Ilya said he was going to let the top go and everybody just thought he was talking about his hair.

[@bdtta6103]

I wanna do coding like that and the computer science stuff like this. This stuff is so interesting. I've also been animating for 4 years now with 2D and 3D work (new YT upload soon). I'm just curious because I am graduating soon, what stuff can I do post secondary related to this stuff. I am also good at math and physics and enjoy doing it (90s in adv function atm, calculus is soon). Thanks man love the content ❤🔥🔥‼️

[@CMak3r]

Let’s hope that Ilya and Sam will come to an agreement on speed/safety of the development. Some compromise is needed to be made because they are both ones of most public key figures in OpenAI. AGI could benefit humanity greatly but risks are also high. Stakes are high and speed of development is rapid, so it’s understandable that disagreements are inevitable

[@IceMetalPunk]

After this, I don't know how Sam could ever trust Ilya again. Instead of trying to come to a compromise, he participated in a coup.

[@agiengineer]

Doctor, It’s not a research Paper Topic😂

[@nufh]

From their blog post, it looks like Sam is a bad guy who chases money over principle. But I saw some key staff resign after they fired him. Great leaders will always have loyal followers.

[@Cyan37]

I suspect its the other way around. From all the interviews I watched, Sam seems to be the careful, slow, do it right kinda guy.

[@Wobbothe3rd]

Sam's whole existence at OpenAI was to make money, he was a business guy not a tech guy lol. That's the paradox at the heart of OpenAI all along! They need a lot of profits to afford to do their nonprofit work!

[@adityapatil325]

Apparently he was trying to open a hardware accelerator business that would've put him in conflict of interest with OpenAI.

[@Nifty-Stuff]

So, this is the company that we're all kinda trusting will be super-duper-careful not to invent something that might wipe out humanity... and then the heads of this company engage in what (from the outside) seems like the most reactive, short-sided, poorly thought-through and terribly execute business decisions at a company this size. Kinda scary.

[@murithiedwin2182]

Great job. However, I must highlight that you changed. About two years ago when I discovered your channel, you used to do "papers". You did real walkthrough of scientific papers of inventions. Fast forward today, your content has become rather somewhat of a news presentation of emerging trends in computer science, and more so AI. I miss when you used to take us through academic papers on new inventions.

[@wisedylan]

I love your quote, what a time to be alive! No other quote is more accurate imo. Thank you for that and everything else you provide!!!

[@nikanj]

Breaking new. He and Brockman both just signed with Microsoft to lead a new internal AI research team there. Microsoft probably paid a pretty penny just to stop them spinning up a serious competitor.

[@MatBat__]

Thanks for the vid!

[@pdcvideostorage7845]

Two Minute Papers’ next video: “Sam Altman joined the evil empire - What a time NOT to be alive!” 😔

[@jerrybot7321]

Battle of "AI" tech bros.

[@PabloBatistaArq]

They're trying to be a monopoly

[@BlueHound]

What a time to be alive!

[@aliticx]

I mean how can the big boys be happy with someone trying to make chat gpt less biased? That’s my conspiracy theory.

[@jojomojo9823]

I suspect people are already working on ways to detect the water mark and remove. In fact.... I wonder if I could ask a certain AI how to write code like that :)

[@yudoball]

Sam altman is fired vsauce: or is he? *music plays

[@jaspreetmaan121]

Sam Altman: My powers have doubled since we last met dooku

[@Bartskol]

What a time to be alive?

[@da_roachdogjr]

Sam Altman is either the hero or the villain. Wish I was a fly on the wall...

[@quenz.goosington]

where's the paper?

[@dreamphoenix]

Thank you.

[@Custodian123]

I'm in the camp to bring strong AI sooner rather than later. Or at least GPT5 level AI asap, at an affordable price. The level of amplification of cognitive work would be never seen before.

[@jamesboone14]

I tried to watch this video to get a good understanding of what's happening, but was unable to make sense of it. The take away here is that there is a lot of speculation regarding the truth.

[@chrissmith4444]

If a not for profit business owns a for profit business. . . .it's for profit.

[@tuseroni6085]

yeah i don't understand how that's legal (i mean..i understand...so corporations can launder money through non-profits and they bribe- i mean give campaign contributions- to politicians who insure they can do that, but i don't know how you can justify that.)

[@Wobbothe3rd]

This.

[@zhang_han]

A non profit company that controls a for profit company. How strange.

[@Wobbothe3rd]

This!!!

[@IceMetalPunk]

It's not strange at all. The nonprofit still needs money to pay for its resources. At scales of billions of dollars required, having a profitable subsidiary is one of the few feasible ways to supplement funding.

[@andrewdunbar828]

If it comes down to keeping either Sam or Ilya but not both, it will stay interesting for a while...

[@LuixAS01]

Wheres the papers 😢

[@ashu-]

It's just a power struggle between Sam and his team pushing AGI and Illya who wants a safety first approach. Open(closed)AI is on the brink of destruction. Sam can't be trusted cuz he is very shady and works for enterprise, Illya is an academic who can't really run a company or be a forefront of one. Opinions are my own.

[@RobShuttleworth]

Some folk's logic is overwritten when large sums of money are involved.

[@AnOligarh]

Safety first is gonna hamper progress, I want to push forward at all costs.

[@celozzip]

good. evil little cockroach. we need less of his tribe in big tech. plus his sister says he used to mess with her when they were kids. should be in jail.

[@AndrewBrownK]

If we can't do it right, is it still worth doing wrong? This is beyond nuclear. This is serious.

[@poetryflynn3712]

@@quicktemperThose people "without heart" know that in time safety constraints will be made regardless of what they do.

[@Mrsparky492]

I love your videos but I think I appreciate that you don't do speculation and industry news normally. The definition of a paper can be stretched to new product releases but personally I don't come here for industry gossip.

[@hydrohasspoken6227]

A non profit organization. Owns a for profit organisation. And they are non profit. Makes a lot of sense, doesn't it.

[@lumbagomason]

When governments get jealous of private companies.

[@colevilleproductions]

imagine if some super spooky scary ai sentience was going on behind the scenes 👁️👄👁️

[@DuxGalt]

Chat GPT has taken over openAI

[@superfoo]

There will be an AI that removes AI watermarks. Key is not to proove generated content but mark genuine content.

[@joech1065]

For sure. I'm not aware of anything in com-sci that will allow you to have a watermark that can't be removed, if that watermark is detectable by other people. The only way to have a non-removable watermark is to make sure that only you can detect it and that for everybody else it appears like random information.

[@IceMetalPunk]

And then you can just add that mark to AI generated stuff, too. Watermarking does nothing in either case.

[@MissesWitch]

really it was the least expected thing to see!!

[@DJ-Illuminate]

Closing the innovation and not allowing it to be public will stop the flow of money and profits.

[@pierce2001]

Did altman want to push open AI into a commercial trillion dollar company and move faster? Or did the board and ilya want that?

[@Cyan37]

Feels more like the latter. Sam had repeatedly stressed the dangers of AI in his interviews.

[@tidelwaver1]

@@Cyan37 After a brief google search I found the opposite. Especially considering that Microsoft, a multibillion dollar for-profit company, is against the resignation. It would make sense for Microsoft to want to push the commercialism of openAI. But I could be easily wrong. I just hope that the AI is developed responsibly

[@jazzwilliams7040]

@@tidelwaver1My counter to that would be that Microsoft would want every other AI to be as open and shared as possible, that way MS could take and use them to improve its own product and inevitably release its own “better” paid product. The advertising would be built in comparability with MS suite of software, as well as customization to any given company’s needs, which many companies would jump at the opportunity to license rather than develop their own home-built solutions that don’t interact with MS (without a fee). Encourage others to be philanthropic to bolster your own ability to turn a profit. That said, I haven’t looked for the interviews - only thinking on what MS corporate interest could be angling at.

[@coryc9040]

He resigned? I thought he was fired?

[@carl00s01]

He was resigned

[@pronobsarker3828]

Made to resign is basically firing with some grace

[@TreesPlease42]

The digital watermarking is so important! 99.99% of people won't bother to remove it, so most AI content will be easily recognizable by platforms -- like recognizing a different image format or audio codec. Ethical AI must use watermarks!

[@DavidBoura]

Until some AI breaks it in 2 nanoseconds.

[@Masterhitman935]

We have better odds with a orbosso AI.

[@fatgnome]

I don't necessarily disagree with watermarking, but as a counterpoint, wouldn't it just create a false sense of security? Especially since anyone who wants to pass of ai generated content as human generated will be incentivized to remove the watermark

[@IceMetalPunk]

If the AI generation isn't mimicking a particular human, and if you can't tell it's AI by listening to it... then why is it an ethical imperative to label it?

[@nicejungle]

1. watermarks are only useful for copyrighting, which is theft, therefore should be banned 2. if an AI can create watermark, an AI can defeat it

[@lobabobloblaw]

I’d sure like to know the full story.

[@cortezforever]

The theory suggests that once OpenAI reached GPT-7, it was prompted to create OpenAI AGI and successfully achieved it. Sam, anticipating his termination, had Greg secure the code through encryption, housing it in a file. The employees found out the next day that they were unable to access the code, which is the motivating factor behind their quest to bring Sam back.

[@protocol6]

So Altman is a superposition of fired and not. Maybe it's a marketing ploy to hype their new quantum GPT. 😛

[@pandoraeeris7860]

Psych! He's back!

[@simlevesque]

Three papers after that, he's back at OpenAI !

[@zurgmuckerberg]

And this company is supposed to 'control' (read: align) AGI when it appears.

[@tank5801]

OpenAI’s CEO was replaced by an AI. 😂😂

[@oshkiv4684]

I think that an honest competitor to chat-gpt would be wonderful for research, and cost for new AI models

[@coco805]

I'll say I'm actually a bit concerned. It could indicate a major change in AI ethics. Now AI is going to pivot to making a profit, and the 'do no evil' could be right out the door again.

[@IceMetalPunk]

It's not mutually exclusive. Building more powerful tech requires resources, which require money, which requires profits.

[@curioplays]

Same worry

[@safakiskyldz9198]

Is the voice here the real voice of him? or an AI or sth? just curios

[@SP-ny1fk]

Ah, here we are seeing the smoke and mirrors that happens when Microsoft suddenly owns AI. If you don't think Microsoft owning AI is an issue - then you deserve what is coming next.

[@humicroav215]

What's coming next?

[@ingusmant]

Watermark is great and all but at the end of the day there are open source local models that can be modified to not have any watermark and state actors of course will not use any watermarks in their AI created propaganda

[@ZacheryGlass]

I thought this channel was about research papers …

[@Wobbothe3rd]

This channel is about AI. Are you really going to argue that this isn't relevant to AI? Cmon.

[@liquidvicinity]

@@Wobbothe3rd hello fellow sort by recent comments explorer

[@mantasr]

Possibly something to do with it not being "Open" AI anymore? Or with the blatant hamstringing of the AI models by the censorship brigade? Who knows...

[@nicejungle]

it's been a while since OpenAI is not open anymore. Before ChatGPT3

[@nefaristo]

I understand the being not so open started 3 or 4 years ago at least, and the censorship brigade is... Pretty much everywhere else, so...🤷‍♂️

[@TheUltraMinebox]

Damn... even Karoly is talking about the coup. Grab your popcorn everyone! Things are about to get interesting

[@realBeltalowda]

This could be fake corporate drams to drum up interest in the company/technology, allowing them to create a new entity or "restructure" and then recapitalize on that fake drama through stock price increases.

[@James-ip1tc]

They're changing the name to DramaAI next week.

[@animanaut]

cant wait for the hollywood movie about it. who should play sam altman?

[@MadeOfParticles]

If AI-generated content is downloadable and recordable, any watermarking method becomes ineffective because it can be removed using a new or existing adversarial neural network methods. There are already excellent open-source options available for AI content creations. The internet will continue to be flooded with this content regardless. Google is trying so hard because they know that an excess of AI-generated content will saturate the internet, causing their advertisers and users to lose even more trust in it.😅

[@tristanwegner]

You repeated the "completely inaudible" water mark too uncritically. Audio compression is DESIGNED, to only leave relevant AUDIBLE data in the file, so I am sure a trained ear/decent audiophil will be able to hear the difference, especially easy in A/B comparisons. It is one thing to believe a scientific paper, that shows benchmarks etc, but a plain marketing statement from a blog? I expect more from two minute PAPERS!

[@TheJmac82]

I would not want to be this board, they cost all the employees of the company millions each. I would be investing in some protection. Also the lawsuits from microsoft investors will be great.

[@_general_error]

That happens when M$ puts their tendrils into a good startup - this is the result of the internal culture, lying of microsoft officials, and endless efforts of microsoft to affect internal politics of other companies. Don't make the same mistake, stay away from Micro$oft and their product, including the $ode editor.

[@thetakburger7928]

Was Sam Altman for slow or quick delivery of products?

[@Eysc]

chatgpt is the board of directors and gives out the decisions

[@DaftCricket]

2 days later and the company is falling apart XD 700+ employees dropped the middle finger in the gc and left 🤣

[@chris83us]

I think they put ChatGPT in charge of the company.

[@Vurt72]

watermarking is absolute garbage. surprised to see anyone if favor of AI to be supportive of that crap. lots of things in here i don't agree with in the slightest. fortunately audio can be passed in an analog pathway and still sound good, doubt a watermark will pass that route, unless its audible.

[@dibbidydoo4318]

This is really odd type of video for your channel, 2MP.

[@Wobbothe3rd]

You should cover the background drama of OpenAI, there has been a lot of interesting institutional and interpersonal conflict before Sam Altman was even CEO, and throughout up to this point. Loon into the Elon Musk threats, etc. I love how Microsoft was caught completely off guard by all this.

[@IceMetalPunk]

If Sam does return to OpenAI, I would be shocked if he lets Ilya stay. This was a full coup, and while the three board members who aren't from OpenAI participated, Ilya's choice here seems particularly traitorous. Especially being done in secret without telling anyone until literally the "you're fired" meeting happens. That's the kind of backstabbing you expect from shady criminal organizations, not one of the biggest tech companies in the world. I don't know how Sam could ever trust him again, so I can't imagine Sam's return would ever happen without a "GTFO, Ilya" clause.

[@OneHappyCrazyPerson]

Faster and more progress! Skynet needs it

[@hydrohasspoken6227]

If i was Sam i would come back and invite those who fired him to keep working in OpenAi. Just to see if they have any honour left.

[@IceMetalPunk]

Just... no longer on the board, right?

[@uku4171]

Let me guess, he got in the way of profits 🙄

[@outspokenwombat2047]

Since this story came out, I am shocked with how naive academics are about the motives behind the people who fund their research. It's time for the researchers of AI and biotech to start paying attention. Your work will be rushed and used in the easiest ways, which are often the worst ways for our society. Read between the lines. Don't put up with this. If that means kicking out the CEO then great.

[@IceMetalPunk]

On the other hand, building new tech requires resources. Where will those resources come from if you "stop putting up with" all your investor's motivations to fund you? It's a balancing act, not a race with one winner.

[@outspokenwombat2047]

Your right, we need a balance. We don't want VC / c-suite people deciding our future. VC always claims its a big rush so they can maintain control. We must decide what is safe, ensure it is, and take our time. Be it AI, genetic engineering, etc. Until we limit corporate and political leaders from using science as a weapon, we can't have nice things. If they can be adults, we have to.