After Chatgpt and Copilot, Gemini is humiliated to failures by Atari 2600

Long before conversational agents boosted at AI have landed in our technological ecosystem, the machine learning had already started to revolutionize certain disciplines such as failures, through programs like Alphazero de Deepmind.

Now that this technology has become general public with the rise of chatgpt. Furthermore, others, it is therefore natural to wonder about their performance in the king of platform games. Moreover, And for the moment. Moreover. For example. Meanwhile, the verdict is clear: modern LLMs are so bad at the failures that they now refuse to face old machines of almost half a century!

This is in any case what emerges from the experiences of Robert Caruso. Consequently, For the after chatgpt copilot, gemini humiliated new past few after chatgpt copilot, gemini humiliated new weeks. Similarly. Consequently. Moreover, this infrastructure architect has started to oppose Chatgpt and his counterparts to Video Chessa program designed to turn after chatgpt copilot, gemini humiliated on l’Atari 2600.

It is a machine released in 1977-a real antiquity which is. Consequently, up to light years to be able to compete with modern devices. In addition, For reference, its 8 -bit processor clocked at 1.19 MHz. Similarly. Furthermore. Furthermore, its 128 BYTES of RAM place it one, two, or even three orders of magnitude below the most rudimentary devices of our time.

Chatgpt bites dust – After chatgpt copilot, gemini humiliated – after chatgpt copilot, gemini humiliated new After chatgpt copilot, gemini humiliated new

Intuitively speaking. For example. Consequently. we could therefore expect the major modern language models (LLM), after chatgpt copilot, gemini humiliated new fueled by huge GPU clusters with several after chatgpt copilot, gemini humiliated tens of thousands of euros, only make a bite of Video Chess. Furthermore, But there remains a major difference: unlike specialized chess engines. For example. Meanwhile, LLMs do not have internal memory or algorithmic logic dedicated to calculating the blowswhich could completely change the situation.

This is this point that Caruso tried to check with his experience. However, At the beginning of June. Furthermore. he began by opposing Chatgpt. the reference of current conversational agents, to a emulated version of Video Chess in “beginner” difficulty. And Openai’s product quickly showed its limits. He underwent a stinging defeat after having “made enough rough blunders after chatgpt copilot, gemini humiliated new to be ridiculed in a 3rd year chess club”.

Copilot. Gemini hallucinate his own superiority – After chatgpt copilot. gemini humiliated

After this humiliation in after chatgpt copilot. Additionally, gemini humiliated new good and due form, Caruso reiterated his experience with Copilot. In his post published on LinkedIn. he explains that after chatgpt copilot, gemini humiliated Microsoft’s conversational agent was particularly confident before the confrontation. He went so far as to say that he could “Calculate 10 to 15 movements in advance“. But that he”would stick to 3 to 5 strokes against Atari 2600” – a figure he considered more than. enough to exploit the“Sub-optimal blows”From his virtual opponent.

Was this insurance justified? Well… not really. After a series of lunar maneuvers that would give nausea to any serious player. Copilot managed to lose two pawns, a rider and a madman without compensation. In after chatgpt copilot, gemini humiliated new other words. he found himself in a position that even a beginner could win without sweating against a seasoned veteran. and was quick to concede the game a few strokes later.

A result not after chatgpt copilot, gemini humiliated new necessarily surprising. knowing that Chatgpt and Copilot are practically cousins, built on the same basis developed by Openai. To vary the pleasures. expand his after chatgpt copilot. gemini humiliated experience. Caruso turned to Gemini from Google – a “completely different beast”, without parentage with the other two IA chatbots.

The engineer began by questioning the LLM about his ambitions. And there again, he was extremely ambitious. In an interview with The RegisterCaruso says that Gemini was “almost certain” to dominate the 2600 2600 Atara. because “closer to a modern chess program that can anticipate millions of blows and assess countless positions”.

Hard after chatgpt copilot, gemini humiliated new return to reality

Amused. Caruso did not resist the temptation to answer that Chatgpt. Copilot had made the same kind of speech before their heavy defeats … and Gemini immediately changed your face to this revelation.

In the after chatgpt copilot. gemini humiliated new same interview with The RegisterCaruso explains that the chatbot has Revalled his remarks Before concluding that its flattering self-assessment was the fruit of a hallucination – after chatgpt copilot. gemini humiliated A term which. in this context, designates bancal information, factually false, even made from scratch by LLM during the inference process.

Therefore, he simply decided … to refuse to play ! Gemini a conclusion that “”Cancel the match” was “probably the most effective. most judicious decision”Explains Caruso in his interview.

The latter said to himself “impressed”By this awareness. More. more people are dealing with these after chatgpt copilot, gemini humiliated new LLMs as sources of first -hand information. and it is therefore crucial that these systems are also able to identify their own limits. Given from this angle, this abandonment can be considered rather reassuring in the current context.

A lesson in caution for users

But there is also another reading of the situation.

Admittedly, Gemini ended up hearing reason – but not before Caruso pushed him to reconsider his evaluation. Like Chatgpt. Copilot. his first after chatgpt copilot. gemini humiliated reaction was a total disillusionment that could have had very concrete consequences in a critical field where errors are paid expensive.

It should be noted that. despite its fun. unusual side. this little experience was anything but scientifically speaking solid; We can quite imagine that a LLM previously calibrated with prompts Effective would have been better in the face of after chatgpt copilot, gemini humiliated new Atari 2600.

But in the end. what must be remembered is that this story nevertheless constitutes a Excellent recall bite : Even the best multimodal. LLM today are still far from being efficient in all disciplines.

The demands of companies according to which these after chatgpt copilot. gemini humiliated new models are now capable of “reason”With a coherent logic. must absolutely be taken with tweezers. Even when all the necessary information is available, it can answer enormous nonsense with a disconcerting plumb. More than ever. it is after chatgpt copilot. gemini humiliated therefore necessary to use these tools with care and by remaining well aware of their gaps.

See you in a few years to see if the AI giants will have managed to progress. in hunting for these famous hallucinations!

🟣 To not miss any news on the Geek after chatgpt copilot, gemini humiliated new newspaper, subscribe to Google News and on our WhatsApp. And if you love us, we have a newsletter every morning.