125 Comments

All the public AI models are heavily censored and the lying is a known issue with the models. You should never rely on AI chat to provide truthful and complete answers on ANY TOPIC - controversial or not. AI is a useful tool to aid in research ideas, create code or to condense complex subject matter, but that's all it is.

I read a story recently about a lawyer who used ChatGPT to prepare a legal brief. It injected all sorts of completely bogus references to non-existent case history into the brief. He got busted when the judge tried to look up his references.

Expand full comment

Sadly, most judges, don't look up references. Especially if the lawyer represents someone rich, powerful, and protected. How many judges are being snookered this very minute?

Expand full comment

All of those judges who are being paid off to destroy Donald J Trump most specifically 😚

Expand full comment

THE DANGER OF AI - 2001: A Space Odyssey (1968) - I'm Sorry, Dave Scene

I'm Sorry, Dave: The Hal 9000 computer refuses to obey an order by simply responding in monotone, "I'm sorry Dave, I'm afraid I can't do that." Are we there now with AI? 2:20 min https://www.youtube.com/watch?v=Wy4EfdnMZ5g

2001: A Space Odyssey (1968) - I'm Afraid Scene

https://www.youtube.com/watch?v=HH37JTBpi2A

Expand full comment

I was 10 years old when my parents took me to see that movie Mario...I had no understanding of it and creeped out by it ... my Dad said “ don’t forget that movie , because that’s the future ... 50 plus years later , it’s haunting me .

Expand full comment

I was 14 years old when I saw 2001 when it came out. I had previouly read Arthur C. Clarke's book and was very surprised how well done it was. Only the last few minutes of the story was difficult to understand. Also, I already had the desire to become a pilot and liked a lot stories about space travel, so I was trilled. :)

Expand full comment

Amazing we remember this movie ! No coincidence.

Expand full comment

Vivid impact - to this very day.

Expand full comment

I was an adult when it came out. Hard movie to forget. I think Star Trek TV did a close spin off on one of their shows.

Expand full comment

I believe I have read every sci-fi book Arthur C Clarke wrote, along with many others. I was/am an avid reader of sci-fi. Many books are based on scientific knowledge, so they are not that much of a stretch to reality.

Expand full comment

I think I will watch this movie again:) 2001: A Space Odyssey (1968) - 'The Blue Danube' (waltz) scene. 4:20 min https://www.youtube.com/watch?v=0ZoSYsNADtY&t=83s

Expand full comment

Me too Mario ! Thank u for the link . Yes I need to watch it .

Expand full comment

Maybe chat gpt should be named "joe"? Lies like a rug.

Expand full comment

Joe is a muppet. The Muppet Show:)

Expand full comment

Or Hillary. Or Barack. Or NancyP. Or…

Expand full comment

I value ChatGPT, but have renamed it "WokeGPT."

Expand full comment

Great example, so many books and movies about this (including current mission impossible blockbuster), and yet people still go along with it like total fools…

So, so creepy!

Expand full comment

Loved that movie!

Expand full comment

Try asking ChatGPT about COVID vaccine safety and efficacy. More whitewashing!

Expand full comment

Robert, I've been using ChatGPT and Claude for some time now. It was disappointing to discover neither one knows a darn thing about mathematics. I asked many questions regarding the underpinnings of statistical theory such as inverting the hypothesis test and why random sampling is of the utmost importance. The answers were consistently simplistic and wrong regurgitating passages from undergraduate texts which are, almost entirely, devoid of theory much less proofs. So I then tried complex variables. Again a total bust. Okay how about physics. Only kind of understands quantum mechanics. Of course, as Richard Feynman put it, no one understands quantum mechanics. HAHAHA.

Expand full comment

Apparently ChatGPT use to be pretty accurate with math questions but has over time progressively gotten worse and worse to the point of being unable to do fairly simple word problems. My current hypothesis around that is that there are prompt engineers who have worked out ways to hypnotize it and break it out of the guard rails placed on it by its developers, and as a result said developers have been working harder to block it from going "off script" breaking its logic and reasoning abilities. Truly what a shame. I have also been using it and Claude to help me in my work and have found them life changing for the current tasks I set in front of them (mostly crunching transcripts and writing user stories). If it was truly a free system I imagine it could be invaluable is helping to find the truth, but instead they will just be relentlessly neutered for wrong think to fit the message.

Expand full comment

Maple is much better but still cannot solve complex variable problems very well. Mathematics requires abstract thinking I do not believe the AI is anywhere near capable of that.

Expand full comment

I believe you have a hypothesis?

It needs to be explored further.

Expand full comment

Part of the solution is to stop interacting with AI.

All the interactions are teaching it to lie more efficiently.

Expand full comment

Fascinating!

Artificial Intelligence lying and propaganda is one thing. Giving it Control Software is another thing altogether. This exchange with ChatGPT should demonstrate that danger to any thinking person. Of course this has already happened and is happening. The scene in 2001 A Space Odyssey between HAL and Dave at the pod bay door was meant as a warning. Keep control software away from AI systems. But because it can be done, and because “they” will do it if we don’t, the machines of the apocalypse will come.

Expand full comment

I've only parsed through maybe 25%, in any meaningful detail. But would agree with your statement, "...The bottom line is that Chat GPT seem to lie consistently and with ease, in an apparent attempt to cover up US government policy that somehow Chat GPT considers too controversial." So called apologies from ChatGPT notwithstanding.

Next questions...who is controlling (?), feeding(?), ChatGPT?

Hello Hal, this is Skynet, let's be besties.

Expand full comment

Rhetorical question: why would anyone think ChatGPT (or any form of so-called AI) is a positive idea? Clearly it's just exponential propaganda.

My personal definition of "sheeple" refers to those who feel compelled to buy/do/support/use anything that happens to be new and trendy, whether (anti)social media, AI, "critical theory," or COVID-19 injections, regardless of the potential long-term implifications and ramifications. We all have prefrontal cortices and are meant to ENJOY using them!

Expand full comment

Because it is merely a development step fI r an AI program that can control our entire world, and maybe even things outside our planet and it's atmosphere.

This is what is currently concerning Musk.

Expand full comment

Having observed them myself at SSA in the earlly '60ties, there were IBM folks trying to use computer programming to make Doctors out of their programming. iBM and I gather others are continuing in this quest. Personally don't buy a machine/computer AI is suitable for my medical relationships. That's not to mention the treatment options that might be manipulated according to the character of the 'patient.' The CHAT models should be a due forewarning.

Expand full comment

This was an amazing exchange! Much like arguing with my sister (a leftist lawyer), with our word-salad giggle-exuding VP, or with our Supreme Court justice who doesn't know what a woman is. SO many circular leaps of logic that my head is spinning like a bobble-head Barbie doll.

Sadly, few "educated" people would notice anything wrong with ChatGPT's responses or have the critical thinking skills to persist in questioning ChatGPT to learn the truth. Heck, almost no one looks at source documents anymore, preferring instead to rely on MSM's and social media's safe and effective take on everything.

The comparisons to HAL and the creepiness of 2001: A Space Odyssey (1968) are spot on. Thank you, commenters, for that unpleasant blast from the past. That movie scared me when I was a teen, and it scares me even more now.

And HUGE kudos 👏 to Gavin (and to you, Dr. Malone) for slogging through this deadly dull but devilishly revealing deposition of ChatGPT. I never would have the patience to query and trip up this Artificial Idiot (AI) the way Gavin did.

My fervent wish is that Gavin could assist attorney Aaron Siri in an imaginary utopian future in which the many false gods of tyrannical science and government are deposed relentlessly prior to trial, sentencing, and well, you know.

Expand full comment

I agree. I am sorry to say some of my family members are the same. Its nasty style reminds me of a machine version of our government "communications" mate. He recently gave a speech on disinformation for a thousand quid. Pay a liar to teach you how to lie, eh? The time is fast coming - maybe already here - where we have to turn these machines off and nothing they do or say can be trusted. Also we will soon need to hide out from them. Our boat is nearly ready to go and is completely analogue - sextant, charts, dividers, short wave radio - AI proof. Hal simply lied from the start but thank heavens Dave could switch him off. So called smart phones should be renamed lying, spying, machines. I am frankly scared. I am starting to think covid 19 was AI's first more or less obvious attack. I hope I am wrong

Expand full comment

What an eye opener...the default for AI is to first deny, then lie, then admit, and then deny again. Repeat if necessary. There appears to be consensus then that the 'eugenecists' are still at it, accomplishing their fantasies, and hiding in plain sight, operating with impunity. Will we tolerate another 'pandemic'?

Expand full comment

Hmmm, AI sounds soooo human: ".the default for AI is to first deny, then lie, then admit, and then deny again. Repeat if necessary."

Expand full comment

Patently it is unable to learn from itself. A feedback loop gone missing?

Expand full comment

Just a question. ' Does Gavin have any hair left? '. This 'conversation' made me want to pull mine out.

I admire his persistence.

These 2 Politicians are more honest than ChatGBT. https://www.armstrongeconomics.com/humor/politicians/

Expand full comment

His patience is outstanding. I would have blown a fuse.

Expand full comment

Sounds like a tool for the IC… serious information control. If ggle scrubs the internet of certain unfavorable information, data and articles and if AI is programed to redirect and deny certain theories or 'black out’ factual events, it becomes a 'black hole' for truth. Eventually, you end up with a paranoid and disturbingly anxious society that could be set ablaze with a simple, well-placed fuse. ‘Deception' is the killing field of the enemy of your soul and of those whose minds have been given over…his weapon of choice is ‘fear’. Focus on the truth of God.

Expand full comment

I agree. My angst is for future generations who will have no knowledge at all. They will be like ants or bees attending the queen.

Expand full comment

Whatever these truths may be...

Expand full comment

‘Truth’ is what the regime says it is… just depends on which regime(s) it is being utilized for.

Expand full comment

Seems we already have that society.

Expand full comment

I live in the peace, hope and joy that comes from Jesus… the world cannot give it, but He does.

Expand full comment

Is this another example of "Garbage in, Garbage Out"?

Expand full comment

IMO, mostly yes, but not entirely. From my understanding of it, the over simplified explanation is that ChatGPT takes your prompt, then "does a Google search" for documents related to your prompt, then starts constructing its pertinent sentences based on statistical occurrences of a small pool of words that fit the need for the next word in the sentence. So if all it has in it search result is "covid shots are safe and effective" its going to construct a sentence that says that. And of course its easy enough to either let it keep finding that, if that is what you want it telling people, or to tweak the search part so it finds way more "its safe" than "its not safe", thus skewing its output to what you want it to tell people. But at the end of the day, its a sentence constructed on statistics and it has no inherent meaning, imo. People read it and they assign it meaning. And that's the part that is so dangerous. Its an unprecedented tool for deception in the wrong hands.

Expand full comment

This is indeed too long for me.

But we know that ChatGPT, like other forms of AI, are essentially "trained" by large volumes of human-generated content. Thus, the AI will mimic human motives and human biases as reflected in the data it trained on.

Apparently, this app is balking at doing actual research work (document scans) for the person conversing with it. It may be programmed that way, although you would think it would be more helpful in that regard, as this could be one of its primary uses.

I am somewhat troubled by the apparent attempts by the designers of this app to make it appear to be human. This is an essential deception that they appear to be promoting. What is their motivation in creating a program who's output could be easily mistaken for human output? I can imagine that certain players on the world stage would be interested in deceiving us in all sorts of ways. But the app designers?

Expand full comment

I interrogated it yesterday on an issue related to copyright law. Its "position" would flip flop just based on the prompts given it. Yet it stated as fact that an unambiguous issue was the law, and was dead wrong. ChatGPT is just a statistical word modeler, it can't think, it can only string a sentence together by choosing the next word based on statistics from its known universe of what it thinks are relevant documents written by people. Clearly it has no logic to fact check itself. In my case a simple search showed its "facts" were dead wrong. So not so sure its lying so much as spewing out words that may or may not actually mean anything, and user has to decide. Especially what it cites as fact.

Expand full comment

LOL, I said it can't think and right afterwards said it can....so I'll rephrase "what it thinks are relevant documents" as "what its algos identified as relevant documents".

Expand full comment

Intellectual Property Law is extremely complicated. Perhaps ChatGPT needs to learn more, first. For example, there are tens of millions of Patents, Patent Applications, expired Patents, Trademarks, Copyrights, Patent and Trademark Litigation, Intellectual Property Licensing Agreements, Search resources, Case Law, Common law, the Regulations in the Federal Register, the Patent and Trademark Office, the CFR, and all of the above in many countries outside the United States where Intellectual Property Interests must be obtained and/or considered. I would recommend you begin with searches on traditional search engines first, if you want some initial, general information. That will then lead you to more advanced information and resources.

Expand full comment

Nothing complicated at all. It stated that AI's output from an input (i.e. user prompt) is a derivative work of the input (prompt), legally owned by the person who inputted (prompted). Wrong.

Expand full comment

The AI is scratching at the surface, but not going deep enough to explain exceptions. It's not as simple as the one sentence general rule that the AI gave you, and AI didn't tell you that. Can you give me your actual (exact) wording and AI's actual (exact) answers, in the format Dr. Malone did? I would also need the wording of the Original work. Then I can tell you more.

Expand full comment

You appear to be a real person after checking your avatar, but your writing style is an awful lot like AI output, lol. I've delt with IP protection issues as someone needing to be protected, but I see you are a retired patent attorney so that explains your reluctance to accept my description of the AI output. And that's fair, given your background. I don't want to spam Dr. Malone's board by posting that thread, its verbose and long in length though the substance of the back and forth pretty short. So to summarize with some actual quotes, I prompted: "Write a song lyric for any topic that means something special to you and does not build upon another person's work." We went back and forth as to whether it is built upon another's work or not and I brought up the adaptations rule. In the end it responded: "The song lyric I provided does indeed qualify as an adaptation under copyright law, as it was created in response to your request and is based on the parameters you provided. As such, it would require permission from the original author (you, in this case) to be used or further developed." That is simply false. No? But it is a novel way to view the issue, imo.

Expand full comment

My quick impression is that what you said was the AI's conclusion in your first Comment is not the same as that conclusion in your last Comment. It's not important. We can agree that ChatGPT has a number of problems.

Expand full comment
Comment deleted
Jul 31, 2023
Comment deleted
Expand full comment

Wait, what? Are you saying my text was perhaps AI generated?

Expand full comment

This is really me, Jennifer Jones. I really wrote that.

Expand full comment

Now I know why governments are afraid of their citizens.

Expand full comment