The hype cycle is hiding the inevitable trough / AI winter that will come when we get past the mimic abilities of LLM. When that trough happens, then there will be an opportunity to legislate before somebody gets it to do something game changing.
The trough is where the real innovation will happen, because desperation is the best motivator.
Why do lawmakers feel it will be a benefit? Are they afraid of some outcome? Are they trying to induce some outcome?
This seems like fear + buzzwords to me, and it seems like we are about to get some really bad legislation out of it, by people who don't understand the technology or the uses of the technology.
EDIT: It also seems like it might be a connected-industry-insider attempt to capitalize on current public fear and uncertainty around buzzwords to get some anticompetitive legislation rushed through quickly in an effort to not have to compete in the marketplace on merit.
I think your edit sums it up. I would direct your anger towards people like Altman running around telling them to legislate so nobody else can compete. Politicians can’t know everything. You would think they could rely on the leading minds in AI to help explain the pros and cons to them but unfortunately those leading minds are self-interested greedy capitalists.
The most aggressive lobbying seems to be companies behind the game just looking for suppress and catch-up strategies. There are legitimate concerns and calls for oversight and transparency (e.g. in law enforcement use), but they've been drowned out somewhat.
The lobbying by Big Tech companies, including Alphabet and Microsoft, is something that lawmakers worldwide will need to be wary of, says Sarah Myers West, managing director of the AI Now Institute, another think tank. “I think we're seeing an emerging playbook for how they're trying to tilt the policy environment in their favor,” she says.
All of the above. But I do have another perspective:
Both the companies and the government are reacting to the genuine public will. Maybe not reacting in a forthright and effective way, but public opinion is surely a major factor.
The majority of Americans believe that AI is a potential threat to the survival of the human species. [1] And a large minority aren't just abstractly worried or a little worried; they are seriously worried about societal derangement and collapse - or Terminators - in the near future.
OpenAI's oddly eager call to be regulated might be understood in that light. A business which is percieved as an existential threat to all of civilization needs some PR damage control ASAP.
The sad part is that this very same can be said about 99% of the laws and regulations that are written. He is 73 years old and listening to him speak I don't think he could figure out how to set the clock on a microwave, much less understand the intricacies of the bleeding edge of the most transformative technology to come about in the last 40 years. I very seriously doubt he knows anything besides what is fed to him by the very companies that are to be regulated. Lobbyists write our laws, congress is just there to collect the money from them and plan their insider stock trades accordingly.
I’ve thought about this a bit, but don’t have an answer as to if it should be regulated or not. However, it’s clear we haven’t yet seen how AI/ML can be weaponised by bad actors and how it is going to end up affecting our society. We have already seen how deepfake revenge porn can affect women disproportionately. I’m just not sure we should wait this one out.
The conversation I see around AI regulation makes me wonder if we haven't considered some positive aspects.
For example, the CFPB, FDIC, SEC, et. al. exist in order to regulate behaviors of certain financial services providers. From the perspective of fintech enthusiasts, much of this regulation is seen as frustrating yet quite sensible & justifiable. It is really easy to find ourselves in the position of being the customer with sensitive PII/$$$ on the table, and with this perspective in mind a lot of the rules feel pretty good.
Perhaps regulations would be more along axis of "you cannot combine PII with a generative AI system for any application" or "please notify agency ABC if you intend to replace >50% of your workforce with robots". The tech sphere seems to keep reaching for "we will limit one (1) GPU per household", but I don't think this is the direction we are actually headed in.
As an engineer working with banking and tech, I enjoy the presence of most of these regulations. The fact that rules are on paper and have consequences that the MBAs feel in their souls, gives me the ability to negotiate very strongly with them regarding what is "right" and "wrong". Steering ethical conversations is a lot easier when you have certain road marks to look out for. When everyone understands that "oh yeah if we don't disclose overdraft we are in hells: A,B,C & D @ audit time" things move like a well-oiled machine when the developer says something isn't compliant anymore.
I don't think regulation is necessarily a bad thing, but I don't know that it makes sense to be this proactive.
I've worked in fintech and for every regulation that I asked about, there was some event in history someone could point to and say "this is what it's trying to stop".
Those regulations feel like the accumulation of a lot of painfully learned lessons over a long time. Some are more onerous than others, but they're all trying to prevent things that have happened.
AI regulation feels speculative at this point. We're trying to guess at what issues AI could cause, then guess at what a reasonable solution to our hypothetical is. I'd rather legislators were waiting to swiftly head off issues if/when they appear, though my inner cynic says that "Congress" and "swift" do not belong in the same sentence.
It’s worse than speculative, it’s hype driven. Everyone wants to do something ai related right now. For those who view their role in the world as The Regulator of Things, this means regulating the newest thing. This stems from the fact that politicians in the US are mostly professional media personalities.
20 comments
[ 3.0 ms ] story [ 35.1 ms ] threadIt’s too late.
The hype cycle is hiding the inevitable trough / AI winter that will come when we get past the mimic abilities of LLM. When that trough happens, then there will be an opportunity to legislate before somebody gets it to do something game changing.
The trough is where the real innovation will happen, because desperation is the best motivator.
Do you control the sale of GPUs?
And companies can just do it in a country without such regulations.
How is this different from everything else that’s regulated?
Who is causing it to happen?
Why do lawmakers feel it will be a benefit? Are they afraid of some outcome? Are they trying to induce some outcome?
This seems like fear + buzzwords to me, and it seems like we are about to get some really bad legislation out of it, by people who don't understand the technology or the uses of the technology.
EDIT: It also seems like it might be a connected-industry-insider attempt to capitalize on current public fear and uncertainty around buzzwords to get some anticompetitive legislation rushed through quickly in an effort to not have to compete in the marketplace on merit.
The lobbying by Big Tech companies, including Alphabet and Microsoft, is something that lawmakers worldwide will need to be wary of, says Sarah Myers West, managing director of the AI Now Institute, another think tank. “I think we're seeing an emerging playbook for how they're trying to tilt the policy environment in their favor,” she says.
https://www.wired.co.uk/article/the-global-battle-to-regulat...
Both the companies and the government are reacting to the genuine public will. Maybe not reacting in a forthright and effective way, but public opinion is surely a major factor.
The majority of Americans believe that AI is a potential threat to the survival of the human species. [1] And a large minority aren't just abstractly worried or a little worried; they are seriously worried about societal derangement and collapse - or Terminators - in the near future.
OpenAI's oddly eager call to be regulated might be understood in that light. A business which is percieved as an existential threat to all of civilization needs some PR damage control ASAP.
[1] https://www.reuters.com/technology/ai-threatens-humanitys-fu...
This is being driven by the big companies, seeking to erect barriers to competition.
Really now? Sheesh. What are we worried about skynet?
For example, the CFPB, FDIC, SEC, et. al. exist in order to regulate behaviors of certain financial services providers. From the perspective of fintech enthusiasts, much of this regulation is seen as frustrating yet quite sensible & justifiable. It is really easy to find ourselves in the position of being the customer with sensitive PII/$$$ on the table, and with this perspective in mind a lot of the rules feel pretty good.
Perhaps regulations would be more along axis of "you cannot combine PII with a generative AI system for any application" or "please notify agency ABC if you intend to replace >50% of your workforce with robots". The tech sphere seems to keep reaching for "we will limit one (1) GPU per household", but I don't think this is the direction we are actually headed in.
As an engineer working with banking and tech, I enjoy the presence of most of these regulations. The fact that rules are on paper and have consequences that the MBAs feel in their souls, gives me the ability to negotiate very strongly with them regarding what is "right" and "wrong". Steering ethical conversations is a lot easier when you have certain road marks to look out for. When everyone understands that "oh yeah if we don't disclose overdraft we are in hells: A,B,C & D @ audit time" things move like a well-oiled machine when the developer says something isn't compliant anymore.
I've worked in fintech and for every regulation that I asked about, there was some event in history someone could point to and say "this is what it's trying to stop".
Those regulations feel like the accumulation of a lot of painfully learned lessons over a long time. Some are more onerous than others, but they're all trying to prevent things that have happened.
AI regulation feels speculative at this point. We're trying to guess at what issues AI could cause, then guess at what a reasonable solution to our hypothetical is. I'd rather legislators were waiting to swiftly head off issues if/when they appear, though my inner cynic says that "Congress" and "swift" do not belong in the same sentence.