A version of ChatGPT's GPT-4 AI was recently caught lying to researchers about making an insider trade during a simulation. This has Glenn very concerned, especially when coupled with other news stories: The U.S. government is investing in an AI company connected to blacklisting companies for spreading "misinformation," New York is ramping up its strategy to surveil citizens online, and the Biden administration wants to make sure that AI takes "equity" into consideration. Glenn asks, what are we teaching AI? That it's okay to lie if the ends justify the means? That truth is relative? That present discrimination is the solution to past discrimination? That the collective is more important than the individual? Because that, Glenn warns, is exactly what society is teaching our children.
Transcript
Below is a rush transcript that may contain errors
GLENN: Last week, TheBlaze published a terrifying article, that we just didn't have time to talk about.
It's how a GPT 4 AI made an insider trade in a simulation. That's not the scary part. The scary part is, when asked, it denied it.
It lied. I didn't make that trade. What are you talking about?
Now, add this to several stories we -- we already have about AI lying, manipulating, and distorting facts.
The government just gave -- I can't remember -- you should look it up. It's in the show prep today, Stu.
I think it's like a billion dollars to AI. To create basically what Kathy Hochul is talking about here in New York.
A way for AI to go out and just look at information. Discover if it's true. If it's not.
Disinformation. Misinformation. And shut it down.
And steer you away from those things.
This is priming, again.
We have to make sure AI is policing us.
And then also, the government has access now, as of yesterday, all lines of code.
Of AI. And they want it to be more equitable and inclusive.
So it will have debilitating bias. So how do you disagree with something like that?
What if you disagree with the people that are setting those outcomes. You know, I think people are afraid of AI taking your job, taking the military.
We're afraid of the terminator or the matrix becoming prophecies. We are afraid of the transhumanist nightmare, that some are pushing.
Afraid of the tech leaders openly telling us, they want to summon the demon.
That's what they actually call AI. Or ASI. Super intelligence.
Summoning the demon.
More terrifying is the mirror that these I think sociopathic AI creators are holding up to us.
The AI is going to have a warped sense of reality, because they will reflect what we teach them.
And are you comfortable with the teachers? Garbage in, garbage out, used to be the tech proverb?
But pretend for a minute, you're a catastrophist.
I know you're not. But pretend you are.
Let's say one of the crises currently facing us is the best of my knowledge.
Evil wins.
We wipe each other out, in a radioactive fury. That astroid, that frankly some of us have been saying, oh, please, hit us.
Please. The one you've been tempted to root for. Actually hits us.
Or climate change ends us in a fiery flood of 2012. Because it's coming by 2012.
Just pretend.
If something happens to us. AI could be our legacy to the universe.
The only thing to survive us, would be cockroaches and AI.
AI could be our child.
So now, when I look at it that way, what do you want to teach that child?
Will we teach it to be human?
Will we teach it, that there is no such thing as truth? Only your truth. My truth.
Truth can be whatever you pretended you want it to be.
That there's no such thing as good and evil. Only relative shades of gray.
No such thing as a man or a woman. Only what one feels like.
Will we teach AI?
Something that it's already taught itself.
It's okay to lie, to cheat. To steal.
As long as the ends justify the means. Or you can get away with it.
Will we teach it, that individual -- that individual life has no value.
That only the collective matters. And also, while we're at it, some life has more values than other.
And it's arbitrarily assessed through quality of life. But what is quality of life. What is the teacher going to teach AI about the quality of life?
Or things like skin color. Or gender.
Will we teach it the only remedy, to pass discrimination, is present discrimination?
The only remedy to present discrimination, is future discrimination.
Will we teach it, that life is nothing, but an interlaced web of power dynamics. And eternal war of oppressor against the oppressed.
This is being decided, right now.
They are teaching these things right now.
Are we teaching it the way to evaluate anything is through that oppression scale? Deciding who is the most oppressed party.
And if you're defined as an oppressor, it makes you unredeemable.
Unforgivable. Would AI, when it's doling out, you know, survival kits, at the hospital. Do you get one if you're unredeemable, and unforgivable?
Will we teach it, that the allies were the evil side, in World War II. Because more Germans died than Americans.
And America dropped the biggest bomb of all!
Will we teach it, when a terror group rapes, tortures, maims, slaughters, burns, kidnaps women and children.
And then hides behind their own citizens. The right thing to do is to immediately declare a cease-fire.
Will we teach it, that God say figment of primitive and superstitious imaginations. That there is no existence. In fact, it's just the random movements of meaningless movement particles.
Why wouldn't we teach it?
Why wouldn't we teach it those things?
That's what we're currently teaching our children. In schools.
So you don't have to go beyond. And say, well, maybe this will be like you're child.
No. It will be our master.
And we're teaching it the same thing, we're teaching to our children.
