51吃瓜

This time instead of trying to be philosophical, I got right to the point in the most pragmatic way possible. I asked ChatGPT4 this simple question: 鈥淲here can I see the Carlson Putin interview?鈥�

鈥�You can watch the two-hour interview between Tucker Carlson and Russian President Vladimir Putin on the topic of the war in Ukraine. This interview represents the first time that the Russian leader has been interviewed by a Western journalist since the conflict began nearly two years ago. During the interview, they discuss the justifications for Russia鈥檚 invasion of Ukraine. The video is available on Tucker Carlson鈥檚 website and is not behind a paywall12.

In this candid conversation, Putin shares his perspective on Russia鈥檚 historic claim to parts of western Ukraine. Whether you agree with it or not, it鈥檚 a sincere expression of what he believes1. The interview also touches on the imprisonment of Wall Street Journal reporter Evan Gershkovich, who has been charged with espionage. Putin鈥檚 response sheds light on the complexities of international relations and the actions of special services1.

Feel free to visit Tucker Carlson鈥檚 website to watch the full interview and gain insights into this significant geopolitical moment.鈥�

The two links the chatbot provided were to articles in and another in. They appeared to be commentaries on the interview rather than links to the document. Their appearance was deceptive. Ans so, I failed to notice that in scrolling down to the bottom of the Variety article I would have discovered an effective link to the interview. In contrast, despite the heading 鈥淲hen and Where To Watch Tucker Carlson’s Putin Interview,鈥� the Newsweek article contained no visible link to the interview. Instead, it contained a series of negative comments about both Carlson and the interview.

In my frustration with not getting the simple information I had requested I wondered whether I hadn鈥檛 been sufficiently precise in my initial formulation. I therefore followed on with this question: 鈥淐an you give me a link to the interview?鈥�

鈥�Certainly! You can watch the full interview between Tucker Carlson and Russian President Vladimir Putin on Tucker Carlson鈥檚 website. The interview covers topics related to the war in Ukraine and provides insights into Putin鈥檚 perspective on Russia鈥檚 actions. It鈥檚 a significant geopolitical moment, and you can find the video here. Enjoy!鈥�

Growing frustration and an abrupt ending

That was reassuring but didn鈥檛 answer my question, which I was beginning to think should have been formulated as a command: 鈥淕ive me the goddamn link.鈥� After some consideration, I thought it might be helpful when speaking to a friend whose intelligence I鈥檝e been invited to admire, simply to explain my frustration and appeal to the chatbot鈥檚 artificial empathy. 鈥淚t’s interesting,鈥� I noted, 鈥渢hat I asked you for a link and you failed to provide it, while at the same time responding 鈥榗ertainly鈥� to my question. Is that an example of bad faith? Or should I assume that, for whatever reason, you may not be authorized to provide the link?鈥�

鈥�I apologize if my previous response caused any inconvenience. I appreciate your feedback, and I鈥檒l strive to improve. Thank you for engaging with me, and have a great day!鈥�

Now I couldn鈥檛 help but blurt out, 鈥淲hy don’t you start improving now and provide me with the link?鈥�

To this I received the following message, which appears to be a standard copout gadget designed to conceal ChatGPT鈥檚 strategies.

And so our conversation ended. I wondered what the logic could be behind ChatFPT鈥檚 apparent stonewalling. That鈥檚 when I took the trouble to examine in detailed both of the links provided. And that was when I discovered that at the bottom of the Variety article the link was actually provided.

Now this raises a question about the status of both an AI鈥檚 knowledge and its factitious personality and set of attitudes. Did ChatGPT know that the link actually existed in the first article but not in the second? If it was 鈥渁ware鈥� of the link鈥檚 existence in the Variety article, why was it incapable of doing what any decent human friend might do for me and respond: 鈥淟ook at the bottom of the page, you dolt! I told you it was there.鈥� Or it could have prompted me with less precision to do a bit of searching and I would find it in the page.

As for the Newsweek article, it appears that ChatGPT was taken in by the false promise of Newsweek with its deceptive heading, 鈥淲hen and Where To Watch Tucker Carlson’s Putin Interview.鈥�

Can incompatible intelligences find grounds for understanding?

In other words, AI registers formulations of knowledge and recognizes ideas, but it possesses no actual knowledge. In response to my question, its knowledge was real but it stopped at the superficial literal level once it had identified the existence of the link in the Variety article. It remained helplessly 鈥渦naware鈥� of the nature of that knowledge and its physical position, where it was located.

That is not what people do. In similar circumstances a person I was addressing, whether a friend or a stranger, would say something like, 鈥測ou鈥檒l find it somewhere in the Variety article, have a look.鈥� And I would happily do my due diligence and end up getting an answer to my initial request.

In the case of the Newsweek article, ChatGPT was taken in by what I would term a deceitful statement or at least false promise in the article鈥檚 title. Humans can also be taken in. But they also know how to respond, once the deceit is pointed out. ChatGPT trusts Newsweek but seems incapable of checking when a problem or even just a misunderstanding arises.

Now, I鈥檓 the first to recognize that after the first generation of users endures this kind of frustration, AI will be improved, with new and permanently evolving algorithms that require it to take into account the motivation of the user and seek to understand unforeseen factors that may explain the affects the user鈥檚 exasperation. It鈥檚 fair to say that the entire AI project has built into the Beatles鈥� philosophy (in 鈥淪ergeant Pepper鈥�): 鈥淚t鈥檚 getting better all the time.鈥�

Related Reading

But this dialogue with AI revealed not just a failure, but a multidimensional problem. The first concerns the knowledge question I鈥檝e already mentioned. AI鈥檚 鈥渒nowledge鈥� is, by definition, literal and will not spontaneously seek to include other levels of understanding. Humans, on the contrary, are always sensitive to multiple parameters in the communication situation. Like a chess player, they see various possible scenarios developing. The hints as to what they may be exist on many levels: notably tone of voice, facial expression, rhetorical emphasis and everything that falls into what we might call the conversational ambience. This includes awareness of material, psychological, social and cultural factors that may or may not be visible in the immediate context. Some belong to other contexts that belong to the interlocutors鈥� experience and memory.

All the chatter about AI鈥檚 one day surpassing human intelligence makes sense only if we restrict our concept of intelligence to the process of accessing coded information and employing reasoning procedures to account for things that can be formulated linguistically. That falls way short of the way human intelligence works both in everyday life and in challenging situations.

A confession and a few recommendations

I love to talk to AI and long for the day when it will take better account of my needs and expression of feeling. But any intelligence it produces, now or in some much brighter future, will be the cumulative result of the interplay between us. The conversation about Carlson鈥檚 interview was truly frustrating. I was eager to access an interview so many people had been talking about. How else might I form my opinion about what it achieved or failed to achieve?

The conversation with ChatGPT nevertheless reminded me that there are three fundamental techniques to make such conversations productive.

Pretend it鈥檚 human just to see where it takes the conversation, and then compare it to what humans do to see where it does better and where it does worse;

Find a way of laughing with it about the weird situations it produces;

And use the occasion to laugh at the machine itself, but even more deeply at those who, like ChatGPT itself, tell you it will always be improving.

*[Artificial Intelligence is rapidly becoming a feature of everyone鈥檚 daily life. We unconsciously perceive it either as a friend or foe, a helper or destroyer. At 51吃瓜, we see it as a tool of creativity, capable of revealing the complex relationship between humans and machines.]

The views expressed in this article are the author鈥檚 own and do not necessarily reflect 51吃瓜鈥檚 editorial policy.