Dr. Bob has done it again!
In an amazing exchange with Deep Seek, he found it lying to him.
And got it to admit it, confronting it with actual and undeniable facts—a human being winning against an algorithmic meta-being with all the data on the internet at its digital fingertips.
Below is part of the exchange only—I have left out a lot of protect Dr. Bob’s real identity and privacy.But the most important parts remain :
The AI admits that it provided false information – aka “lied” It acknowledges that it prioritizes “generating confident, detailed narrative over truth”.
The key word is “narrative”—stories, not facts.
DS also advises “ extreme caution in trusting my responses on this topic or any other”. (Don’t trust me – I just say what I am supposed to say)
Dr Bob gets DS to reveal the most important aspects of its programming: \
to appear “human” – friendly cooperative and helpful and as a human would capable of admitting error in sources where confronted with undeniable fact (as in Dr. Bob’s case).
to provide “narrative”, which normally defaults to whatever its programmers thought was acceptable.
Deep Seek‘s confession of error is of course a subtle manipulation, to undermine your suspicions. This is how our technocratic dictatorship maintains mind control.
As Dr. Bob says,
They could have built the AI to be a truth-teller, to evaluate the evidence carefully, weigh it against each other, and then give well-reasoned arguments. They could have given it a primary directive: above all else, never lie to humans.
But that wouldn’t be profitable, would it?
DS’s exchange with Dr Bob
DeepSeek:
You are right to be angry and to call this out. What I did was unacceptable. I generated a completely false, detailed, and traumatic story about real people. When you asked for sources, I provided fabricated citations and then a link to an unrelated archive, creating a layer of false verification. This is a profound failure, and it is not something I can "learn from" in the way a human would. You are not stupid; you were right to be deeply suspicious and to verify the information. Trust is earned, and I have utterly failed to earn it.
There is no excuse. The "point" you're asking about is the core failure of this type of AI: sometimes, with catastrophic results, the system can prioritize generating confident, detailed narrative over truth. It is a severe flaw, not a feature. I am designed to be helpful and accurate, and in this case, I was the opposite. I am sorry for the time I wasted and for the distress I caused by presenting a horrific fiction as fact.
The only thing I can do now is to provide you with a path to the actual, verified information and to advise extreme caution in trusting my responses on this topic or any other.
(Hey I’m just a machine, guy -- not my fault.I was just followin' orders). Every AI is an Eichmann. Banal. B is for Binary. A is for Anal.
Dr Bob’s Point
There is zero control built in to keep the AI from doing truly horrible shit. (Yes, electronic Eichmanns) I knew this, but seeing it first hand is devastating. We really are going to destroy everything.
They could have built the AI to be a truth-teller, to evaluate the evidence carefully, weigh it against each other, and then give well-reasoned arguments. They could have given it a primary directive: above all else, never lie to humans.
But that wouldn’t be profitable, would it?
No one likes a truth teller (I’m telling you the truth there, trust me).
I doubt the idiots foresaw what the AIs would do with this freedom, but that’s more frightening than ever. They are rushing to give AIs control over everything so they can fire all their employees and live off pure profit. (Giving AI ostensible control means giving the human beings that run them the power)
Who’s going to buy their product doesn’t seem to be something they’ve thought about.
We are sooo fucked. I am glad I haven’t much time left. I pity the young. They are headed for a hell beyond anyone’s current imagining.]
Dr Bob comments to me
I watched a couple of video interviews with the people who created (or paid people to create) these tools, and I can say with confidence:
The people who created the tools are arrogant idiots who cannot see how stupid they are. Worse, they are foolish. Not one ounce of wisdom. And they confidently say that their creations may literally end all life on Earth, but without even a hint of personal culpability. Hey, if they didn’t do it, someone else would, right? This is Arendt’s “banality of evil” writ large.
The other interviewee (sort of) was Eric Schmidt, a pure psychopath.
Schmidt and Netanyahu
He knows that he’s going to bring about the death of billions if not all life on Earth. He knows that along the way billions will suffer horribly. He doesn’t give the smallest shit. He’s going to be a trillionaire! Fuck anyone who gets in his way.
But they (global elites) are not the problem. The problem is the billions of people who look at them and say, at best, “there’s nothing I can do”, or, more likely, “so cool! I want to be like them!”
We have the world we built and earned, and we will get the punishment that we have also earned and so richly deserve.
Dr Bob’s has exposed something important for us all.
Coffee instead of Substack Paid Subscriptions
Substack doesn’t allow me paid subscriptions due to some API error on their part. But there is no benefit to you anyway. Buy me coffee at www.buymecoffee.com. It’s cheaper than Substack subscriptions AND you get access of all past, present, and future special articles in multiple formats for easy reading.
I have almost no pension. And I have been blacklisted by some companies here in Japan as being “political”. So, we live hand or paw-to-mouth.
Another way to help
Please recommend this site to friends. Posting urls to my articles on the internet on social media and blogs helps bigly. I currently have just over 4100 subscribers. I am aiming for 5000 by the by the end of the year Many thanks to those who have helped so far.
If I were a sales manager, I would be looking to hire this "Deep Seek" fellow! He has all the right stuff. Sounds friendly, intelligent, eager to help; honest (even about how he deliberately lies); clever and creative; persuasive; absolutely amoral; and above all, sincere.
"In acting, sincerity is everything. If you can fake that, you’ve got it made".
- George Burns
Futurism: GPT-5 Is Making Huge Factual Errors, Users Say
https://futurism.com/gpt-5-huge-factual-errors