Discussion

• @benmschmidt @benmschmidt on x

  I think we can call it shut on ‘Open’ AI: the 98 page paper introducing GPT-4 proudly declares that they're disclosing *nothing* about the contents of their training set. https://twitter.com/...

• @jjvincent James Vincent on x

  OpenAI has shared very few details about its latest AI model GPT-4, triggering criticisms from many in the AI community. Co-founder and chief scientist Ilya Sutskever told The Verge that the company's past approach to sharing research was “wrong” https://www.theverge.com/... http…

• @elonmusk Elon Musk on x

  @Carnage4Life I'm still confused as to how a non-profit to which I donated ~$100M somehow became a $30B market cap for-profit. If this is legal, why doesn't everyone do it?

• @stokel @stokel on x

  GPT-4 looks great - but the fact that OpenAI won't say anything about how it works is a worry for many experts. My latest for @FastCompany, featuring @CatherineFlick, @technollama, @IEthics and @npseaver https://www.fastcompany.com/ ...

• @alexjc @alexjc on x

  Interesting take by @technollama that OpenAI's secrecy around GPT-4 is a legal defense. Since it's a multimedia model, it could well be! Also, if it's true they're keeping things secret for legal reasons, it's a sign they know they're non-compliant and have no good defense ;-) ht…

• @austen Austen Allred on x

  Ah, it's a thing https://twitter.com/...

• @emostaque Emad on x

  Open offer to anyone @OpenAI who actually wants to work on Open AI: We will match your salary, benefits etc but you can work on any open source AI projects you like, ours or others. Collaborate, be open and prioritise good outcomes over self interest: https://openai.com/... https…

• @jeffladish Jeffrey Ladish on x

  Great to see people able to change their mind and own it, especially about really consequential stuff! I hope OpenAI will change their mind about scaling at max speed, especially because the reason it's so dangerous is the same reason that wide open publishing is so dangerous htt…

• @tshevl Toby on x

  The original “open” AI branding was a mistake. Elon was wrong, and still is wrong. Very few people who appreciate the risks of AGI think AGI labs should be sharing everything

• @__aston__ Aston Motes on x

  Common misunderstanding! OpenAI is actually just a temporarily embarrassed nonprofit currently operating as a for-profit until they get their act together. In the end they'll be a nonprofit again. https://twitter.com/...

• @amuse @amuse on x

  It is vital that @HouseGOP hold hearings on AI. How did OpenAI take hundreds of millions in donations to build a nonprofit and open AI to selling out to Microsoft for billions? Get to work @Jim_Jordan we need answers. AI will change America, don't we get a say in how? https://twi…

• @emostaque Emad on x

  You will also get all the compute you can eat, frens, community and the ability to work in awesome tropical locations. Plus respect https://twitter.com/...

• @jjvincent James Vincent on x

  Some have suggested one reason OpenAI does not disclose training data for its LLM is potential legal liability, as with the lawsuits against AI image generators like Stable Diffusion. I asked about this too: https://www.theverge.com/... https://twitter.com/...

• @mervenoyann @mervenoyann on x

  I mean not explaining one thing you don't do and expecting others to understand why you don't do it is a bit (????) also why do we keep using the phrase AGI so conveniently is a bit weird as well https://twitter.com/...

• @cherthedev Cher Scarlett on x

  I'm not saying transparency is always good, but privatizing this will not stop the harm. It will only make it more difficult for the rest of us to anticipate it and defend against it. https://twitter.com/...

• @janusrose Janus Rose on x

  really funny to have a company called “OpenAI” which uses datasets that are not open to train automated systems that are neither artificial nor intelligent https://twitter.com/...

• @bengoertzel Ben Goertzel on x

  Nice to see OpenAI being open about their abandonment of openness. But their justifications are specious. Yes vigorous openness would be bad for Microsoft's biz model. But to make like the safety advantages of closed centralized control of AGI are obvious.. um @SingularityNET htt…

• @bcmerchant Brian Merchant on x

  We need to talk more about how OpenAI uses the looming power & malevolence of Artificial General Intelligence—that it says it's creating!—as a marketing tool. Altman is constantly doing this: “no one can be trusted with this tech, but us” It certainly makes people want to try it!

• @warrenisdead @warrenisdead on x

  If you believe, as we do, that at some point, puppet shows are going to be extremely, unbelievably potent, then it just does not make sense to let people see your hands. https://twitter.com/...

• @tshevl Toby on x

  Agree with @ilyasut on this. I was actually pleased to see that @OpenAI didn't reveal details of how they trained GPT-4. https://twitter.com/...

• @arthurb @arthurb on x

  This will likely invite responses filled with cynicism, in part because asking “qui bono” is a cheap way to sound smart, but the simpler explanation is that they finally realised something that should have been obvious from the start and corrected their trajectory. https://twitte…

• @unormal @unormal on x

  hm yeah corporations doing this entirely black box definitely seems like the way to definitely have nothing bad happen because corporations acting unmonitored are definitely incentivized to do good things correctly and safely history has shown this!! https://twitter.com/...

• @shashidigital Shashi Shekhar Vempati on x

  Serious question given the exponential pace of AI evolution and the implications to societies across the globe https://twitter.com/...

• @marie_haynes Dr. Marie Haynes on x

  Here is an important thread to read. We don't know much at all about the training set for GPT-4. Given that many of the problems arising from AI chat come from issues with what the data was trained on this could be a huge concern. https://twitter.com/...

• @carnage4life Dare Obasanjo on x

  Whenever you call Sam Altman to ask why OpenAI is no longer a non-profit or even “open” https://twitter.com/...

• @esyudkowsky Eliezer Yudkowsky on x

  Actually encouraging that anyone who works there is able to say oops out loud, though (a) OpenAI needed to say that much earlier and (b) they needed to change their name much much earlier. https://twitter.com/...

• @bcmerchant Brian Merchant on x

  “Our GPT-4 AI model is so potent that it would be wrong to open source it—it's simply too powerful. Infusing the technology into a tech giant's search product for $10 billion and commoditizing it for personal profit—now that is a-ok” https://twitter.com/...

• @buccocapital BuccoCapital Guy on x

  Very interesting plot twist if Altman Musked Musk https://twitter.com/...

• @rahelsuess @rahelsuess on x

  Not surprised by this transition. GPT-4 is the most secretive release yet from the formerly nonprofit research lab, now fully transitioned to a for-profit tech firm. #GPT4 https://www.technologyreview.com/ ...

• @jesswhittles Jess Whittlestone on x

  I'm quoted in this article by @jjvincent on whether OpenAI should have released more details about how they built GPT-4: https://www.theverge.com/... Some more detailed thoughts below:

• @joonian Wong Joon Ian on x

  the crazy thing is thiel was right... crypto is libertarian and AI is communist https://twitter.com/...

• @bcmerchant Brian Merchant on x

  In the early days OpenAI was a “nonprofit” researching the hard questions of AI and keeping its “dangerous” models private for the good of humanity—now the whole enterprise seems like a calculated effort to gin up science fictional hype-levels around the eventual products

• @grady_booch Grady Booch on x

  Trust us, says @OpenAI https://twitter.com/... https://twitter.com/...

• @jjvincent James Vincent on x

  debates over over vs closed research are becoming more pressing, as tech giants like Microsoft and Google accelerate AI integration (and sideline ethics concerns. See: microsoft laying off a team dedicated to ensuring AI products meet ethical standards)

• @ewindisch Erica Windisch on x

  These LLMs are directly strip-mining open source software. AI platforms will happily take your mundane code and remix it into commercial and military applications. https://twitter.com/...

• @floydmarinescu Floyd Marinescu on x

  GPT-4 is another reminder that we must share the gains from artificial intelligence as broadly as possible. The people at OpenAI know this. Their governing nonprofit is structured so that it can “sponsor the world's most comprehensive UBI experiment.” 1/7 ⤵️ https://twitter.com/.…

• @austen Austen Allred on x

  I still don't *fully* understand how OpenAI turned from being a nonprofit that had raised a ton of money from donors into a for-profit company or that you could simply do that

• @vogon Yves Adele Fartlow on x

  it is simply irresponsible to allow people to audit the machine that we want to hand control of all of human society over to https://twitter.com/...

• @iethics @iethics on x

  Open-sourcing is not the same as disclosing what #data a model was trained on (which is not the same as making that data accessible to all), and the claim that “training data is #technology” needs to be challenged : https://theverge.com/... #ethics #AI #business h/t @mmitchell_ai

• @1br0wn Ian Brown on x

  The idea there will #AGI for many, many years is just... total 💩 and certainly not a justification for hiding the source of training data for #GPT4 https://twitter.com/...