But that happens with humans as well. You are having the same experience with an AI that many managers have with their direct reports.
The smarter AI gets, the easier it becomes to move the AGI goalposts. Seems at this point there are people who will refuse to call anything less than omniintelligence AGI.
(And then the excuse will be, but it’s not omniscient! And even if it were, is it omnipotent?)
My success rate for solving software engineering challenges encountered in my day jobs has been near 100% for my entire career. I can only think of a few tasks I kicked back and said they were impossible. For example, after trying to get a signal processing system working reliably I decided to sit down and calculate the actual limits of the channel we were sending the data over and found that from a basic estimation it would not be possible to do. In start ups you don't really get to get stuck in a spiral and not fix things.
I find agents often get into these cases during research tasks.
Yep, people are typing comments with a computer that is powered by several layers of software that will be stored on another computer powered by several layers of software to be read on a computer also powered by layer of software. And then they hope to make the argument that humans cannot produce software.
> * Continue tons of hype about how good they are without delivering, going to great lengths to publish how their model "hacked" its way out of a sandbox they misconfigured.
That wasn’t Anthropic. Clearly not a well informed take.
No need. There are already plenty of them that track the S&P — which notably decided NOT to change its rules.
Feel like many of these articles were either prewritten, lazy, or have intentionally omitted the non-impact that S&P opting to not change its rules has had, just so the headline and lede could be as sensational as possible.
It’s frustrating to see the top commenters all trying to provide original takes, while only demonstrating they did not read the article which already makes *the exact same points!*
Seriously, it’s a good article. Read it. And yes, it explicitly discusses ICE.
> where you explicitly tell the visa officers you're on US soil ONLY for study purposes
What in the world does “ONLY for study purposes” mean? 24 hours a day, every day of the week?
> participate in a protest against the very country that granted you the study visa and then get mad that you are under investigation and would have been kicked out for violating the said visa? That's so bizarre.
First, he briefly attended the protest. Not the same as participating. I doubt the data from Google indicated he was holding a sign, shouting slogans, or speaking on stage. And it doesn’t sound like there was any marching or sit-in involved. (And if so, for 5 minutes?)
Second, why are you willfully equating a pro-Palestinian protest with being an anti-US protesT? Was the purpose of the protest to raise charitable funds, encourage more open discussion about the war on campus, provide moral support to Palestinian classmates, and/or any of a myriad of other purposes?
Finally, even if the purpose of the protest was politically motivated —- to push US policy on Israel and Palestine to change, how is that bizarre? In your mind is any protest that seeks to change a government’s policy at that moment an assault on that government, or on that nation? Someone who protests the death penalty, protests for stronger/weaker abortion laws, stronger/weaker gun laws, etc?
This is the USA we’re talking about. Despite all our faults (and they are legion), it is the bedrock of our founding and our core principles that democracy is a participatory process. Not just on Election Day. Throughout history we have advanced as a people and a nation because individuals have stepped up and spoken up. That has always been what has pushed us forward.
>First, he briefly attended the protest. Not the same as participating. I doubt the data from Google indicated he was holding a sign, shouting slogans, or speaking on stage. And it doesn’t sound like there was any marching or sit-in involved. (And if so, for 5 minutes?)
You misunderstand. I'm not against protesting, nor am I against the reasons behind his protests. He may have had valid reasons. What I'm saying is - if you are a green card holder or a citizen, this would be very little risk vs going to a foreign country in a study visa and doing what he did. If you pay tens of thousands of dollars to get a degree from whatever country, for whatever reasons, why would you want to gamble all of it?
Also, if you are getting into a fight, you need to make sure you have the upper hand. As it stands, it is him who is in hiding and crossing borders, not the government agents or the corporate white collars that gave away his data. That's my point.
Sorry I was a bit unclear - this was referencing the Administration not Anthropic.
The United States Department of Justice under the Trump administration, supported Phillips.[20][5] While the Department asserts that anti-discrimination laws are necessary to prevent businesses that provide goods and services from discriminating, these laws cannot be used to compel a business into expressing speech they do not agree with, nor used to provide goods and services with such expressions without the ability for the business to assert they do not agree with those expressions. [1]
> these laws cannot be used to compel a business into expressing speech they do not agree with
I don’t think they are preventing them. You can still buy some Anthropic. Heck, maybe you’re now doubly motivated to buy some out of spite. What Uncle Sam is saying is he doesn’t want anything to do with it. But Anthropic can produce whatever and others can buy it.
In their view just like businesses cannot be compelled so are the customers, they can’t be compelled to buy.
If they're designated a "supply chain risk", then any company that does any business with the military cannot be a customer. That includes basically all the largest companies, many of which have already adopted Claude. So the Pentagon is threatening Anthropic with terminating most of their private enterprise revenue and basically ruining their business model.
That's a little different than just denying someone government contracts.
> If they're designated a "supply chain risk", then any company that does any business with the military cannot be a customer.
Wrong.
Companies with military contracts cannot rely on Anthropic-supplied products and services for those contracts. (Yes, the cabinet member who misrepresents his own title and name of his department also publicly misrepresented the legal consequences of the designation. It's almost like ignoring the law and just making things up is a pattern with him and his boss.)
If you were a customer, what would you do? Keep paying for Claude but be extra careful about preventing all the people working on anything that might potentially be construed as related to the DoD work from using it, for fear of a retributive Hegseth? Or just use codex company-wide and not worry?
> The designation, typically reserved for foreign firms with ties to U.S. adversaries, could ban companies that work with the government from partnering with Anthropic.
> “You’re telling everyone else who supplies to the DOD you cannot use Anthropic’s models, while also saying that the DOD must use Anthropic’s models.”
I wouldn't put anything past this administration in terms of twisting rules and acting in bad faith to excerpt as much leverage as possible. They do it all the time. It's basically a government of patent trolls threatening everyone with meritless lawsuits that are nonetheless extremely effective.
If I was “one of the largest companies”, as was raised upthread as being impacted in all of their business, then I would be used to having many large public and private customers with different and conflicting contracting requirements and segregating support for those contracts, and for US defense contracts specifically, probably have a dedicated business unit for those that probably a subsidiary legal entity and which, in any case, is almost completely walled off in practice dedicate to defense contracts, provides all the shared services consumed by individual defense contracts independently of the parent corporation, and which adheres strictly to defense contracting rules and charges the compliance costs back to those defense contracts at a healthy profit, while having basically no impact on how the rest of the company does business.
You can imagine all sorts of hypothetical scenarios where Anthropic doesn't suffer too much. You can also imagine them losing a lot of big business. The point is that the DoD is sending a very clear signal: "if you don't do what we say, we will punish you until you do". If they didn't want to punish anthropic, they would simply go to a competitor like OAI. The fact that they're threatening several different potential revenge plots proves otherwise.
The govt has so many levers it could pull that it's technically not allowed to but that this administration has made very clear it loves doing. Things like spurious lawsuits prosecuted by a perpetually unconfirmed AG, or capriciously interfering in mergers or permitting processes. There's not a single norm too far for these guys. You're Dario Amodei. You would not be comforted by the idea that they're "not allowed" to punish you.
> If they're designated a "supply chain risk", then any company that does any business with the military cannot be a customer.
That's not how it works, they'll just have to show how they mitigated the "risk". That doesn't mean not doing business with Anthropic that may mean not using Anthropic for any deliverables or any projects involving the specific contracts.
> they'll just have to show how they mitigated the "risk"
If Hegseth is the one who decides whether the risk has been mitigated (he is), you think he's gonna be overcome by a sudden spirit of good faith and make impartial judgements? Or just do the thing that maximizes his leverage, gratifies his ego, and pleases his boss.
> Or just do the thing that maximizes his leverage, gratifies his ego, and pleases his boss.
It doesn't really work that way. Both parties want something from each other. If he is not "overcome by a sudden good faith" judgement all of the sudden no more Windows updates and it's RHEL Linux for everyone. Or if IBM says no, then what? Write your own OS? The system doesn't really work as a charity, it's corrupt but parties want something from each other. If he knows they need something and there is no other way to get the spirit of "good faith" will descend like lightning upon him. In this case he knew there is Google and OpenAI in play, and just like magic OpenAI made a deal pretty quickly.
It's normal to simply go to a competitor when one supplier isn't giving you what you want. It's not normal to try to ruin their business relationships with everyone else in retribution.
You know the only reason we used double spaces between sentences in the typewriter days was because everything was monospaced, right? Modern variable width fonts provide additional space automatically.
The smarter AI gets, the easier it becomes to move the AGI goalposts. Seems at this point there are people who will refuse to call anything less than omniintelligence AGI.
(And then the excuse will be, but it’s not omniscient! And even if it were, is it omnipotent?)
reply