This fix is in!!!
Anthropic and OpenAI have bought protection from @realDonaldTrump — it seems!
Insane that after they both stole IP they’re looking to shut down access the free open source models that stole from them
Regulatory capture worked — and they did it in plain sight!!!
5% of their stock and $100m in lobbyists!
THIS ISNT A DEMOCRACY!!!
It's always amazing to me how intelligent and thoughtful individuals become disrespectful and closed mined when talking about politics.
Politics is toxic, partisans are insufferable.
The more you know, the more you realize how much you don’t know. The less you know, the more you think you know everything. Knowledge is humbling. Ignorance is arrogant.
If an AI tells you to kill yourself, will you just believe it?
No, because if you are an AI First thinker, you check out EVERYTHING it tells you.
At least the important stuff.
If a doctor tells you you have cancer, don't you get a second opinion? Yes you do.
Last weekend I saw software teams building this kind of validator that will check out all facts that, say, GPT reports. But it isn't a solved problem yet, so we must check out everything.
Was just having a discussion with someone who was upset that it "lied" to him.
No, you discovered a bug.
Congratulations, there are many bugs. Many many many.
The machine is new. It has not been debugged fully.
But then neither was Microsoft Windows. Microsoft shipped with many bugs still in the product.
(We had real time bug chart on the wall when I worked there).
It is only a machine. It isn't human. It isn't accountable.
That said, it helps me think, even when it lies.
And there are ways to fix the bugs.
Ask it to debug its own code. Point out its bullshit. Sometimes it apologizes and gives you a better answer. Sometimes it apologizes and lays down new bullshit.
Validate everything.
Oh, and you must do this with ALL digital communications. The technology is now there for, say, your son to call you on the phone and convince you that he's in trouble.
He might be, but you must validate that. You might be talking to an AI and not your son at all. And the AIs can be convincing and confident.
Validate everything. Get a second opinion on everything.
Hey @BrianRoemmele when you build me an LLM make it validate everything, even multiple times and multiple ways. Then add a social layer on top so we humans can discuss answers with others.
Humans are amazing validators. A good second opinion actually is to ask four other smart people. Could an AI call me one day and ask me to validate something? It sure could. And I'd have to validate it. So much validation.
Is there a unicorn in AI validation? I think so.
We all will want more validators in our lives soon.
In 1794, Eli Whitney invented the cotton gin, a machine that drastically sped up the production of cotton
The cotton gin didn't end slavery.
Without first solving society's issues, AI won't make the world better
It'll just *efficiently* make the world worse