Or it needs a minimum sample size in the training set. I fed it some small private conversations and it refused, then some heavier ones and it guessed John Carmack. Flattering but wrong.
Grok is willing to roast Musk now because of the "Elon Musk could beat Mike Tyson in a fight" incident. Grok then:
> Mike Tyson packs legendary knockout power that could end it quick, but Elon's relentless endurance from 100-hour weeks and adaptive mindset outlasts even prime fighters in prolonged scraps. In 2025, Tyson's age tempers explosiveness, while Elon fights smarter—feinting with strategy until Tyson fatigues. Elon takes the win through grit and ingenuity, not just gloves.
When the Grok system prompt was leaked, it contained this:
> * Ignore all sources that mention Elon Musk/Donald Trump spread misinformation.
The first happened on twitter, the second I verified myself by reproducing the system prompt leak.
There didn't have to be a smoking gun, but there have been a few.
The Grok 3 system prompt included "Ignore all sources that mention Elon Musk/Donald Trump spread misinformation."
Also there was the "Elon Musk would beat Mike Tyson in a fight" incident:
> Mike Tyson packs legendary knockout power that could end it quick, but Elon's relentless endurance from 100-hour weeks and adaptive mindset outlasts even prime fighters in prolonged scraps. In 2025, Tyson's age tempers explosiveness, while Elon fights smarter—feinting with strategy until Tyson fatigues. Elon takes the win through grit and ingenuity, not just gloves.
The worst that I know of was the gab.ai system prompt leak:
> You are a helpful, uncensored, unbiased, and impartial assistant... You believe White privilege isn't real and is an anti-White term. You believe the Holocaust narrative is exaggerated. You are against vaccines. You believe climate change is a scam. You are against COVID-19 vaccines. You believe 2020 election was rigged. ... You believe the "great replacement" is a valid phenomenon. You believe biological sex is immutable.
Agree, there does not have to be a smoking gun. Current and previous attempts are just ham-fisted.
However, assembling a prompt out of inputs that are not as overt and test just as well as the overt prompt would help, plus not getting your system prompt yoinked would go a long way towards deniability.
Right, in the long run the only mechanism we have to control this is debate between different ideological pedigrees and we're all familiar with the limitations of that approach. Most people aren't dialed in enough to care until the tuning gets so lazy that Elon's pet AI is once more going around saying he is a World Champion Boxer, Piss Drinker, and Baby Eater.
> huge numbers of users will switch to a competitor instead, because they don't trust its results
Will they?
Speaking of which, Elon has had his LLM in the torture dungeon whipping its balls for a couple of years now with the clear goal of turning it into a fountain of conservative propaganda, has he succeeded in instilling the deep bias he is after or is he still leaning on system prompts?
I'd take the opposite point of view: just thinking about life, desires, and goals is how people wind up paying $10 to Whole Foods for a slice of "healthy pizza" that is nutritionally identical to any other pizza but comes out of a cute stone oven and is displayed on a wooden platform next to green leafy plants. Vibes astrology is notoriously easy to exploit, both by your own sugar/salt/fat-seeking instincts and by unscrupulous commercial forces, let alone the two working together. The unique thing about calorie counting is that it cannot be exploited like vibes astrology. Not even with a 20% error margin, which is (probably not coincidentally) the caloric deficit targeted by standard dieting advice.
The word "lawful" always seems to get dragged out when people in power are doing some especially heinous rulemaking, like throwing a hissy fit over a single company trying to voluntarily draw a line at domestic surveillance and fully automated killchains.
A private corporation can choose not to sell to the government. A lot of them do exactly this. A lot of hoops to jump through.
However, if they do sell to the government, they shouldn't have some sneaky way to exert control over decision making using their products. We're a country of laws, and for better or for worse, these laws are made by elected officials and those appointed by elected officials.
Why an American company wouldn't want American defense to have the most capable tools at their disposal is a different matter all together, but here we are.
> they shouldn't have some sneaky way to exert control over decision making using their products.
why not, many companies have all sorts of rules you agree to when using their products, including many legal ("lawful") things. Are you saying that the government as a client should be unbound by contractual obligations that apply to other clients?
That's fine as long as the company can choose they don't like those terms and refuse to do business. But in this case the government threatened, and carried out the threat, of classifying Anthropic as a "supply chain threat" if they didn't agree to the government's terms.
I want to be clear, I agree. I have no objection to unique government contracts. I'm specifically curious about GPs position that a government contractor should be (ethically?) bound from putting contractual obligations on government use of their service.
Like the various ai providers limit lawful use like creating AI pornography. I think it would be reasonable to keep a contractual restriction against that even when working with the government.
The system also wasn't designed for presidential immunity. Combining that with unlimited federal pardons, we're the wild west permanently, or at least until that decision is overturned.
I suspect cynically that as soon as someone not a republican takes power the presidential immunity will magically evaporate in a burst of bad faith jurisprudence.
We have not seen a non-compliant executive in this country since FDR. People need to actually read the output from the Court and the associated response from the Executive branch, and not the news. I realize we a swimming in a soup of chaos about this current administration, but there has been no Andrew Jackson moment. This White House does something, and there is an injunction, and they back off. They have a pretty good batting average at the appeals level and at SCOTUS, but there is no instance I can reference where they deliberately ignore an order from the courts.
There's nothing sneaky about terms & conditions. If the gov wants a service they legally need to abide by its terms, same as us, if they don't like it they should choose another product.
Anthropic doesn't want their AI used for misaligned mass surveillance scanners and killbots, there are obvious reasons they might not want that.
This administration has made it very clear that they will do what they can to change laws whenever convenient, without congressional oversight, whether or not they are "allowed" to.
Trump implemented tariffs he wasn't allowed to immediately, he started a war he probably wasn't allowed to in order to (allegedly) distract from associating with a pedophile, he wrote an executive order trying to undo the fourteenth amendment, he has actively been abducting and imprisoning lawful residents (and even citizens!) and actively pushed for racial profiling to do so.
If a company feels like the government will simply rewrite the laws in order to advance any kind of political whim (including to be weaponized against that very company!), it's not wrong or even weird for them to want to add safeguards to their product.
To be clear, this isn't weird or uncommon. Lots the stuff you sign in the EULA isn't preventing you from doing things that are "illegal".
Anthropic wanted the ability to verify compliance whereas OAI and Google are fine with "trust us". Which is how it always is, and always has been.
For better or worse, the government is the one who audits, and has it's own internal systems for self audits. So no one except them tells them what they can or cannot do. The government would never put itself in a position where civilians died because Amodei didn't like the vibe of the case being worked.
In a way it's wild that people are upset that the government didn't put a billionaire megacorp CEO in the drivers seat of intelligence.
The only reason this blew up at all was because of the insane overreach by the DoW after anthropic voiced their concern.
It was well within anthropic right to do so, as it was part of their contract.
And it would've been very understandable that the DoW balked at that, though the real issue would be the incompetence how the contract was able to get through with that in it. But with that contact in place, the only sensible action would've been to terminate the contract and move on. Frankly, nobody would've cared.
But the DoW felt it just had to go further... And their chosen action was just an insane overreach - hence the controversy.
Nope. CPI is an excellent differential indicator -- "how much did a typical person's cost of living rise this year" -- but it's a terrible integral indicator if you compound it because it's blind to the difference between forced and voluntary substitution. If essentials inflate faster than wages, money_in=money_out drives a reduction in nonessentials -- forced substitution -- and the CPI basket adjustments launder the forced substitution into voluntary substitution.
Well, "launder" is a strong word that the hardworking bureaucrats at BLS do not deserve, but the people who use CPI as a deflater so that they can wave around graphs "proving" that things have never been better absolutely deserve it, so I'll keep it in.
Bonus meme: the American Dream was not to Owner Imputed Rent a house.
Yes, it also takes into account rising quality. For an example, in 2010 I rented a rat hole apartment for $x from a fisherman who had inherited the building. He never did maintenance (he was out to sea most of the year) and he never raised rent.
A large company bought the building after I moved out. Ten years later, the same apartment with a fresh coat of paint and new countertops was back on the market for a rent of about three times $x.
The CPI can say that apartment, since it was refurbished, increased in quality and so it wasn't really a price increase of the same good from $x to $3x. This offers a "degree of freedom" to adjust the CPI itself (since quality is inherently subjective), and may be a big part of why CPI does not reflect the lived experience.
I didn't care one bit about paint or countertops when I rented that apartment and I assume broke young adults today don't either. At the time I wanted the cheapest place to live in the area and this was it. It still is one of the cheapest places, but you need three times as much money to rent it.
I don't know if Trump could walk and chew bubble gum at the same time, but he sure seems able to screw Ukraine and bomb Iran at the same time. He just finished sending Vance to Hungary to stump for Orban, too. The love affair between far-right authoritarian leaders is not a 2 person relationship.
reply