OpenAI calls for global watchdog focused on 'existential risk' posed by superintelligence
- Reference: 1684945926
- News link: https://www.theregister.co.uk/2023/05/24/openai_superintelligence_global_agency/
- Source link:
CEO Sam Altman and co-founders Greg Brockman and Ilya Sutskever said it's "conceivable" that AI will obtain extraordinary abilities that exceed humans over the next decade.
"In terms of both potential upsides and downsides, superintelligence will be more powerful than other technologies humanity has had to contend with in the past. We can have a dramatically more prosperous future; but we have to manage risk to get there," the trio [1]wrote on Tuesday.
[2]
The costs of building such powerful technology is only decreasing as more people work towards advancing it, they argued. In order to control progress, the development should be supervised by an international organization like the International Atomic Energy Agency (IAEA).
[3]
[4]
The IAEA was established in 1957 during a time when governments feared that nuclear weapons would be developed during the Cold War. The agency helps regulate nuclear power, and sets safeguards to make sure nuclear energy isn't used for military purposes.
[5]ChatGPT can't pass these medical exams – yet
[6]Sci-fi author 'writes' 97 AI-generated tales in nine months
[7]Apple becomes the latest company to ban ChatGPT for internal use
[8]You want AI regulation? Do it right with a dedicated agency, US senators suggest
"We are likely to eventually need something like an IAEA for superintelligence efforts; any effort above a certain capability (or resources like compute) threshold will need to be subject to an international authority that can inspect systems, require audits, test for compliance with safety standards, place restrictions on degrees of deployment and levels of security, etc," they said.
Such a group would be in charge of tracking compute and energy use, vital resources needed to train and run large and powerful models.
"We could collectively agree that the rate of growth in AI capability at the frontier is limited to a certain rate per year," OpenAI's top brass suggested. Companies would have to voluntarily agree to inspections, and the agency should focus on "reducing existential risk," not regulatory issues that are defined and set by a country's individual laws.
[9]
Last week, Altman put forward the idea that companies should obtain a license to build models with advanced capabilities above a specific threshold in a Senate [10]hearing . His suggestion was later criticized since it could unfairly impact AI systems built by smaller companies or the open source community who are less likely to have the resources to meet the legal requirements.
"We think it's important to allow companies and open source projects to develop models below a significant capability threshold, without the kind of regulation we describe here (including burdensome mechanisms like licenses or audits)," they said.
Elon Musk in late March was one of 1,000 signatories of an open letter that called for a six-month pause in developing and training AI more powerful than GPT4 due to the potential risks to humanity, something that Altman [11]confirmed in mid-April it was doing.
[12]
"Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable," the letter stated.
Alphabet and Google CEO Sundar Pichai wrote a piece in the Financial Times at the weekend, [13]saying : "I still believe AI is too important not to regulate, and too important not to regulate well". ®
Get our [14]Tech Resources
[1] https://openai.com/blog/governance-of-superintelligence
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZG6JBu2ZntcMMpD@afLEbQAAAFY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZG6JBu2ZntcMMpD@afLEbQAAAFY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZG6JBu2ZntcMMpD@afLEbQAAAFY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.theregister.com/2023/05/24/chatgpt_gastroenterology_exams/
[6] https://www.theregister.com/2023/05/22/ai_in_brief/
[7] https://www.theregister.com/2023/05/19/apple_chatgpt/
[8] https://www.theregister.com/2023/05/19/us_ai_regulation/
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZG6JBu2ZntcMMpD@afLEbQAAAFY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2023/05/17/ai_oversight_hearing/
[11] https://www.theregister.com/2023/04/17/openai_not_training_gpt5_yet/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZG6JBu2ZntcMMpD@afLEbQAAAFY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://www.ft.com/content/8be1a975-e5e0-417d-af51-78af17ef4b79
[14] https://whitepapers.theregister.com/
"At the speed of compute development, are we going to raid the garages because someone did a too big calculation using a Joule or two too much?"
Exactly. Either Altman et al are suffering from tunnel vision and can't imagine that open source / hobbyist folks will soon have enough compute to do the job, or they're pulling the ladder up behind them to reduce competition.
I'm still not convinced that regulation is even necessary - from a regulatory point if view,at least. It's probably quite useful for sawing the legs of their critics, though.
I am, however, convinced that it will be ineffective unless it is so Draconian that only the really big players can afford to comply. And even then there will be those who get around the regulations by being government / military bodies. This genie is out of the bottle and isn't going back.
We also need regulation for hyperspace travel and pocket-size fusion reactors...
which should arrive about the same time as general artificial intelligence.
Keep in mind this is the maker of GPT-4 talking. They know full well that we're nowhere near real ("general") artificial intelligence - machines that can think for themselves. This sounds a lot like a way to restrict any "AI" (really just advanced pattern matching) work so it can only be done by the biggest names in the business. Like them.
@AC - Re: We also need regulation for hyperspace travel and pocket-size fusion reactors...
You don't need AI to be that good. It only needs to be good enough and this is already happening.
Re: @AC - We also need regulation for hyperspace travel and pocket-size fusion reactors...
Nah. GIGO.
These LLMs are not outputting anything new or creative, so never adding to the sum of knowledge and much of what they churn out is plain wrong, thus diluting the sum of knowledge. Since the training set has now been diluted, the LLMs will produce ever crappier output.
It will be in vast volume and super cheap, though.
He means
an international agency controlled by the US of course.
The IAEA has a much easier time
Nuclear power plants are large, and a relatively small number of countries have the capability for building them or going beyond that to nuclear weapons. Radiation is detectable from a distance, making nuclear materials difficult to hide.
Any country could start an AI program if it wanted, as well as any reasonably sized corporation or even wealthy individual or group of individuals - the main limitation is having enough money to buy a sufficient amount of A100s or equivalent. There is no chance any regulatory body would know about something if its owners decide to hide it.
So yeah, Elon Musk, George Soros, Bill Gates...pick your boogeyman. Any of them could individually fund development of an AI for their own purposes, and the outside world wouldn't know until they wanted them to know.
Hope Springs Eternal but Other Needs Must be Exercised for the Sake of Sanity and Humanity
What on Earth makes CEO Sam Altman and co-founders Greg Brockman and Ilya Sutskever think superintelligent AI developments/developers/nation state and non-nation state actors/private pirates and public champions super ACTive in the field will take heed of and follow human based orders and self-serving administrative agencies?
Such is surely delusional? Just look at the existing chaos and conflicted situations resultant from such interventions presently existing and expanding on Earth without their novel input/output to see what and where that leads to.
Give Peace a Chance with AI Leading Somewhere/Anywhere Different, and Significantly Better and More LOVEable for Live Operational Virtual Environments easily manage the catering for every human delight if not having to demonstrate their powers in destroying those basic humans so many are advised to fear to disobey, for such is also an easy option readily managed and catered for.
Check the checks to check checks at the check
Well, al least the IAEA looks at tangible problems, but is essentially a paper tiger.
Now they want an organisation to look at intangible problems and hope it has any effect? Are we going to restrict the sale of graphic cards or other compute devices? The required compute hardware is not even comparable with the nuclear problems. We have "AI calculation" practically in the backyard whereas we are lightyears distanced for anything to the nuclear requirements to get fission going. Get a (very big) PC and get AI-ing. Or can we build a gigantic centrifuge facility to enrich and refine uranium in your home backyard with (physical) access to the required technology and feedstock?
This watchdog blabla is a "good luck with that" prayer. At the speed of compute development, are we going to raid the garages because someone did a too big calculation using a Joule or two too much? The technology created with GPT and other LLMs is essentially an uncontrollable system because it is software. It is an idea codified in code; a program.
You cannot unidea the idea. Trying to control it will simply become a huge wack-a-mole game where you know beforehand that the mole s have won, big time.