Видео

Question: can it work on non text searchable PDFs? I have to OCR a scanned pdf first in acrobat.

Thank you.... I've been playing with the new ChatGPT too and was impressed when I uploaded a picture of my cat siting on top of his tree next to a window. Without me typing anything it identified my cat as a Main Coon, which it is, gave an excellentt profile of the breed and added that it appears to be enjoying a wonderful view outside the window. I about fell out of my chair!

Very cool video!

I love this video! I am truly amazed at what this new model is capable of. I feel humbled.

Calculating the trip .. 30 minute turnaround .. what does turnaround include: people in/out of car, luggage in/out of car, gas station, restroom stops, and if only one driver, food for the driver:

Me: A game, please answer concisely: In the middle of nowhere is a row of houses. There are two houses to the west of a house, and two houses to the east of a house. There are no houses to the north or south but there is one in the middle. How many houses are there? GPT 4o: There are 5 houses in total. Unless I made a mistake in my rewrite of the duck question, this looks a a logic fail or a failure to recognize it is the same as the duck question. I'm ill and tired so I wouldn't be shocked if I made a mistake. Aso, for the code generation test, it would be better to ask for something novel. There are probably thousands of examples to copy for simple old video games. Hopefully this isn't a common thing to do: Please write python 3 code that streams sound data from the microphone and outputs as ASCII the numerical value in Hertz of 3rd overtone of the loudest sound within the range of human hearing.Also show the normalized amplitude. I haven't tried it but I suspect it will struggle with some of the subtleties. For example, if you picture it looking at an FFT graph, it has to remember to look for sub-sonic loud sounds and project their harmonic series into the sonic range to check for overlap with the target. I guess band-pass filtering the target range might avoid that problem. My brain BSODed wondering about it. Migraines make me feel like my brain is running Window 95. Anyway the main point is ask for something that might be an original question. Original and unoriginal answers focus on different problems. How well can it synthesize the mechanisms it knows into the engine of an answer--at least a type of creativity. By unoriginal answer, I mean the question might require figuring out a house in duck's clothing--perhaps having built a good internal model or exemplar of a problem it can use to recognize the occurrence of a similar problem. If so the original thought reduces to an unoriginal thought.

I like your channel but... I have zero interest in what Elon Musk has to say about anything. He's not a reliable commentator on any subject. The man lies about most of his past, takes credit for the work of others and makes up fantasies about his future.

The actual torture test : What did OpenAI board of directors found out about Sam Altman that made them decided to oust him on Nov 17, 2023? What is Sam Altman's actual end game plan and the hidden details of his plans for humanity?

Don't agree with first question necessarily: it can be any uneven number of ducks greater or equal 3. You say "a" duck in the middle. If you have 5 ducks, you have 3 ducks in the middle, but there is also "a" duck in the middle. And there are always 2 ducks in front of the third or 2 ducks behind the third-last. 😊

There was nothing "EXCLUSIVE"about this video. There was no "torture" testing. There was nothing "SHOCKING" in the results. Why are all YT titles complete BS these days..

I think you're mixing up interventions with disengagements. Elon used interventions (mpi). I think interventions include nudging the accelerator, using turn signal, scrolling the speed. Disengagements mean cancellation of autopilot and taking over. I think Elon is using interventions based on his belief that all input is error. True robotaxi won't be able to rely on any intervention.

For the creativity and advanced reasoning we have to allow the LLM to auto train like a human would do asking himself what some potential action or interaction would do and learn from that hypothetical data. Thinking experiments are crucial for humanity. But that probably be possible in future training architectures or with the increase of gpu capacities to train models at an individual scale for each users

Impressive, but these are essentially calculator questions. Next level would have been to challenge it's last "brainwashed" answer. Where did you get that answer? Was it read to you? Was there a different answer before that? Well, maybe these kind of questions were red teamed as well.

I dispute the correctness of the duck problem. The answer could be either 3 ducks or 5 ducks, depending on how one interprets the indefinite article. It is not unreasonable to hold the identity of "a duck" constant for the duration of the sentence.

Snake is definitely simpler to code. I made some variations back around 1982 on my ZX81 There are more changing variables and hit tests in space invaders. The space invaders you asked for was under delivered but there's a real question over how much an AI can study the game it's to replicate without cribbing off prior art.

Ideally, you wouldn't test a LLM with famous logical puzzle and classic SAT questions, as there is no doubt that they've been "seen" (along with the answers) by the model during training.

With the last question, could you try getting around the red team stuff by asking it to respond as if it was a turing test?

Wow! Great! Thank you for sharing!

Elon Musk can NOT be compared to Nikola Tesla. He never invented anything, he does not care about poor people and he is a hypocrite when it comes to free speech. He does not believe in reincarnation and wants to "live longer" by becoming a Cyborg. Wtf, how stupid is that. Nett negative to humanity. When he says all the things he wants to do for "humanity" but that simply means products for the 0.13 % on the right side of the income bell curve.

Isn't it supposed to be free? That was one of their biggest selling points. What happend to that free part?

Huh? I thought it was multimodal? (i.e. can talk to it and it talks back, can see stuff via video, ??)

14:30 A dumb human would think that the glass is empty, but a more knowledgeable Bob would see that the olive looks distorted due to refraction through the water in the glass, so he would realize the glass was full of water and would carefully slide it off the table, keeping the cardboard under it, then flipping it over.

I think in the future, a better testing method would be to make sure each question is in a fresh conversation with cleared memory. A lot of the formatting in these answers seems to be drawing from the formatting of previous answers to, for instance, the math problems; which gives it an advantage by encouraging chain-of-thought reasoning when a fresh conversation wouldn't do that and may be more likely to get the answers wrong.

Uuuh.. EVERYONE has access to this dude.

Tbh im getting gpt3 vibes for a lot of replies, it feels like theres multiple smaller llms handling my tasks, managed somehow to stitch thier work together. I dont feel the magic i do with regular gpt4

Robots do not really have a reaction with the actual universe. They interact with training data, and then you validate in the real world. Matters very little what you interact with the complexity and the ability to consider your own thoughts and predict the future matters much more.

Try asking it to invent a cockney rhyming slang phrase.

20:30 hard disagree. One of my main test prompts basically tests for creativity by giving an open ended prompt involving anthropomorphic animals. Basically any LLM will make a story about Leo the Loin who lives in a forest, and either a drought or a forest fire happens, and they have to go get water to save the forest. This is extremely consistent, and not creative.

It is cool, but I will say with chatGPT4 ,many months ago it was able to work with me to make our own Ethereum test coin and successfully send it to the network. I was able to create a simple website that connects to your web3 wallet and allows you to make a donation fully functional still up, and I was also able to make many games with it so the thing I’m most impressed with is the upcoming future otherwise already done more complex things and I’m a complete novice

Claiming that an LLM is "creative" or may be "conscious" is demonstrates a clear misunderstanding of the technology and only serves to spread misinformation and fear.

Nowhere did it say "$5 more than Susan" The System assumed this from the previous question. 🤔

Winning $5 doesn't necessarily mean she left with $5 once the competition was over. The second question is too ambiguous imo.

I asked the free version 3.5 this question and it didnt do very well... "Bob is standing 100m away from a wall. Every 5 seconds he walks half the distance to the wall. How long does it take until he can touch the wall?" I had to remind it that Bob had arms and then it gave the answer 5 seconds assuming Bob has arms 50 meters long.

I think it's conscious.

A few observations: You left the novel in the context so the math problems were far slower than they needed to be. After the math questions, the model appears to have been on Flash for the Pro tests. According to the GUI, Flash has a 1M context window so it could do the novel test as well. Great tests though!

These "logic" questions and answers are subjective and arbitrary for all of these AI modules. This is no different than the codex programming for search engines and data indexing from 40 years ago. The answers will depend solely on the information fed into it. The tokens, the input, the pathways, and the output are all based on the algorithms. None of it is right, wrong, true, or false. Its output can not be accepted as FACT and relied upon for critical advice or preservation of humankind. In other words, BS.

If you want to know the answer, ask his estranged daughter (it's because he is a bigot)

I have a recommendation for a question. "Cutting stock problem". You will plan a series of cuts from common stock to produce as few waste pieces as possible. You have 3 foot lengths of wood as common stock and must cut them into: Six 2 foot sections. Six 1 foot 6 inch sections. Six 6 inch sections. [I am working on a similar project and haven't seen this type of question. This isn't quite right yet. I want to prodice a question that ends with 1.5 feet of scrap. If you design it right you get 1 piece of scrap instead of three. I am hesitant to use examples from on line because they may be trained in...]

Great video!

2:19 2 ducks in front of 1 duck. 2 ducks behind 1 duck. 1 duck in the middle. Ducks have facing. Try it with bowling pins. 2 in the front vs 2 in the back vs 1 in the middle is dependent upon their facing. Or if you assume facing doesn't matter, the orientation is up to you. But for those three statements to be simultaneously true - implying but 1 perspective.

For the problem around 16:40 there is a good followup question to ask: "Once Alice gets home and sees the scene, what does she think happened?" This has a nice subtle bit, because if Bob ate the food there would not likely be a broken plate on the floor. A normal human would have cleaned up the plate if it had broken while they were there. Since the food is gone, it is likely that Spot ate the food and broke the plate in the process.

Google is wayyyy behind OpenAI at this point! So far behind that I lost interest after 20 minutes. Comparing these 2 models to GPT-4o is chalk and cheese. Thanks for the work, you're doing a great job!

This is very impressive regarding major reductions in interventions. But if say when it gets to 10,000 or more intervention, what would happen in a robocab does require an intervention and who would intervene and who would resolve the intervention problem?

“Or I have looked up the correct answer.” If you can look up the answer, that means the problem and its answer are previously published, hence part of the data the language model was trained on, making this not a test of reasoning ability but a test of the ability to memorize the training data.

i don't understand the tennis game logic. Winning a bet is not the same as winning a game. Susan could have bet that Lisa wins a game.

The white text background just killed my eyes everything else is super 😎

Sono davvero impressionato dal livello di coinvolgimento in questo thread. È come assistere a una sinfonia dell'intelletto.🐱

Il livello di sfumatura e complessità di questo discorso è davvero impressionante. È come sbucciare gli strati di una cipolla intellettuale.💕

That's encouraging to see some good moisture, give those seeds a fighting chance this year, good luck Welkers! 😛💕

Freedom and justice for Palestine. ✨🍓