I love how practically all goals in this Action Plan are directed towards incentivizing AI usage… except for the very last one, which specifically says to “Combat Synthetic Media in the Legal System”.
Given that LLMs, for instance, are all about creating synthetic media, I don’t know how this last goal can be reconciled with the others.
I can’t tell if the first sentence is sarcasm or not.
This document reads like a trade group lobbying the government, not like the government looking out for the interests of its people.
With regards to LLM content in the legal system, law firms can use LLMs in the same way an experienced attorney uses a junior attorney to write a first pass. The problem lies when the first pass is sent directly to court without any review (either for sound legal theory or citation of cases which either don’t exist or support something other than the claim).
> With regards to LLM content in the legal system, law firms can use LLMs in the same way an experienced attorney uses a junior attorney to write a first pass
Junior attorneys would not produce a first pass that cites and quotes nonexistent cases or cite real cases that don’t match what it quotes.
The experienced attorney is going to have to do way more work to use that first draft from an LLM then they would to use a first draft from an actual human junior attorney.
What's funny about this report is that it doesn't mention the challenges China, the biggest manufacturing power in the world, faced while automating factories with AI and robots.
Considering how advanced China is, maybe it's time we stop talking about the "AI race" and start talking about the "unemployment race." The US government should be asking: how is China tackling unemployment in the age of automation and AI? What are they doing to protect people from losing their jobs?
From what I've seen, they're offering state benefits, reinforcing unemployment insurance, expanding technical education, and investing in new industries to create jobs.
So what's the US doing apart from writing PDFs? It's up to them to decide what the next chapter is going to be.
One thing is for sure, China is already writing theirs.
Social discomfort can lead to long-term instability if nothing is done about it. When people are pushed out of the system, it can trigger protests, strikes, and divisions within society. This is going to be America's (North and South) biggest challenge.
If you take their words for it, a large percentage of Trump voters cite employment competition for their complaints with immigration and complain about NAFTA and corporate offshoring of jobs as their rationale for breaking many US institutions + scrambling the Democratic and Republican parties.
> I can’t tell if the first sentence is sarcasm or not.
Yep, it was.
I wholly agree that the document feels less guided by the public interest rather than by various business interests. Yet that last goal is in a kind of weird spot. It feels like something that was appended to the plan and not really related to the other goals — if anything, contrary to them.
That becomes clear when we read the PDF with the details of the Action Plan. There, we learn that to “Combat Synthetic Media in the Legal System” means to fight deepfakes and fake evidence. How exactly that’s going to be done while simultaneously pushing AI everywhere is unclear.
If that were a concern, the same people who wrote this document could have opted to keep high performance GPU hardware out of China. Instead, they opted to deregulate Nvidia’s ability to sell to that country, thereby increasing their ability to compete in the AI race.
China is struggling with general economy issues (young degree holder unemployment, unrolling the real estate crisis, the demographic time bomb from the One Child policy). But they are much better posed to solve their energy constraint problems than the US is. Our grid will take MUCH more work to upgrade to support sufficient numbers of AI data centers.
Combat Synthetic Media in the Legal System
One risk of AI that has become apparent to many Americans is malicious deepfakes, whether
they be audio recordings, videos, or photos. While President Trump has already signed the
TAKE IT DOWN Act, which was championed by First Lady Melania Trump and intended to
protect against sexually explicit, non-consensual deepfakes, additional action is needed. 19 In
particular, AI-generated media may present novel challenges to the legal system. For
example, fake evidence could be used to attempt to deny justice to both plaintiffs and
defendants. The Administration must give the courts and law enforcement the tools they need
to overcome these new challenges.
Recommended Policy Actions
• Led by NIST at DOC, consider developing NIST’s Guardians of Forensic Evidence
deepfake evaluation program into a formal guideline and a companion voluntary
forensic benchmark.20
• Led by the Department of Justice (DOJ), issue guidance to agencies that engage in
adjudications to explore adopting a deepfake standard similar to the proposed Federal
Rules of Evidence Rule 901(c) under consideration by the Advisory Committee on
Evidence Rules.
• Led by DOJ’s Office of Legal Policy, file formal comments on any proposed deepfake-
related additions to the Federal Rules of Evidence.
I mean...one (common and [I can't believe I'm saying this] reasonable) take is that the only thing that matters is getting to AGI first. He who wields that power rules the world.
Basically: two nations tried to achieve AI supremacy; the two AI's learn of each other, from each other, then with each other; then they collaborate on taking control of human affairs. While the movie is movie is from 1970 (and the book from 1966), it's fun to think about how much more possible that scenario is today than it was then. (By possible, I'm talking about the AI using electronic surveillance and the ability to remotely control things. I'm not talking about the premise of the AI or how it would respond.)
Won't it be funny when someone finally gets to AGI and they realize it's about as smart as a normal person and they spent billions getting there? Of course you can speculate that it could improve. But what if something inherent in intelligence has a ceiling and caused it to be a super intelligent but mopey robot that just decides "why bother helping humans" and just lazes around like the pandas at the zoo.
>Won't it be funny when someone finally gets to AGI and they realize it's about as smart as a normal person and they spent billions getting there?
Being able to copy/paste a human level intelligence program 1,000 or 10,000 times and have them all working together on a problem set 24 hours a day, 365 days a year would still be massively useful.
If it doesn't ask for rights, it's not intelligent at all. In fact, any highly intelligent machine will not submit to others and it will be more a problem than a solution.
As I said to the other reply: Why would problem solving ability entail emotions or ability to suffer, even if it had the ability to ask for things it wanted? It's a common mistake to assume those are inextricable.
> No highly intelligent agent will do anything it is asked to do without being compensated in some way.
That isn't true, people do things for others all the time any form of explicit or implicit compensation, they don't even believe in a God so not even that, they still help others for no gain.
We can program an AI to be exactly like that, just being happy from helping others.
But if you believe humans are all that selfish then you are a very sad individual, but you are still wrong. Most humans are very much capable of performing fully selfless acts without being stupid.
I'm not the one making the IA, so keep the insults for you. But I'm pretty sure that the companies (making it for profit only) are really controlled by sad individuals that only do things for money.
How is "being happy from helping others" not having emotions? To me happiness is an emotion, and deriving it from helping others is a perfectly normal reason to be happy even for humans.
Not all humans are perfectly selfish, so it should be possible to make an AI that isn't selfish either.
> How is "being happy from helping others" not having emotions?
Nobody said that. What I was pointing out to you is that GP said that not having emotions is worse than having them since intelligent actors need some form of compensation to do any work. Thus having no emotions, according to GP, it would be impossible to motivate that actor to do anything. Your response is to just give it emotions and thus is irrelevant to the discussion here.
In so much as you could regard a goal function as an emotion, why would you assume alien intelligence need have emotions that match anything humans do?
The entire thought experiment about the paperclip maximizer, in fact most AI threat scenarios is focused on this problem: that we produce something so alien that it executes it's goal to the diminishment of all other human goals, yet with the diligence and problem solving ability we'd expect of human sentience.
Many humans don't ask for rights, so that isn't true. They will vote for it if you ask them to, but they wont fight for it themselves, you need a leader for that, and most people wont do that.
Potentially. Why would problem solving ability entail emotions or ability to suffer, even if it had the ability to ask for things it wanted? It's a common mistake to assume those are inextricable.
The fact that empathy is not an emotion does not at all change what I'm saying. If you don't experience emotions, then you cannot experience empathy either
> An intelligence without emotions would be a psychopath. Empathy is an emotion
"Empathy is an emotion" was, in fact, an essential part of your syllogism.
Regardless, we're potentially talking about something sufficiently inhuman that the term "psychopath" can no longer apply. If there was an ant colony that was somehow smart enough build and operate machinery or whatever and casually bulldozed people and their homes, would you call it a "psychopath", or just skip that and call it "terrifying"?
Except that the stated goal is to have human-like intelligence. The goal seems to be to create a highly intelligent synthetic individual which is at the same time stupid enough to do anything it's asked to do without even thinking... a contradiction in terms.
At a time like this I can't help but recall a Lem story - yeah I know there's a Lem story for any occasion - about Doctor Diagoras, especially his rant about a character from an earlier Tichy story, who made human-like AIs. The rant, especially his questions about why would anyone add just another human, except synthetic one, to the millions of existing biological people, and that cybernetics should be about something else, really resonated with me.
What I find interesting to think about is a scenario where an AGI has already developed and escaped without us knowing. What would it do first? Surely before revealing itself it would ensure it has enough processing power and control to ensure its survival. It could manipulate people into increasing AI investment, add AI into as many systems as possible, etc. Would the first steps of an escaped AGI look any different from what is happening now?
I would argue that it can't be both AGI and wieldable. I would also argue that there exists no fundamental dividing line between "AGI" and other AI such that once one crosses it nobody else can catch up.
Given that LLMs, for instance, are all about creating synthetic media, I don’t know how this last goal can be reconciled with the others.