Correction for the sponsor spot at 1:02 – We meant to say "14700 KF", as is shown on screen 🙂

🤦‍♂using musk as a source for the future of AI

We are just meat computers ourselves anyway reacting to stimuli through our senses and programmed to do whatever is in our interest either personally or collectively at any given time. Consciousness/sentience is just the means of processing said stimuli in my view, so applying this logic that it is this higher level, almost supernatural thing doesn't really make sense.

You can only talk about current AI.
And I can make current AI do things you would not think possible.
Current AI is the worst it will ever be. And before you are done making up your mind on current AI, it's moved on and become even more capable.

Atm ai is just just a decen pattern recognition algorithm

don't wanna hate but this is untrue, i use it for software engineering and it accelerates many tasks like boiler plate code of database modelisation etc, So why were you hyped by AI features in your other video " Hey Siri, I’m Giving You A Second Chance ", AI is not a lie it is useful and i use it myself to accelerate my game deveopment for complex features that would take a hour in only a few seconds + testing ai results and one shot success is happening even more with AI complexity i only worry about the environment for its computational cost (energy consumption at a world scale)

All these AI chips being added to PCs and phones scanning what your doing are there for one reason. They are there for spying and the governments will love this. If you're not scared of Apple Intelligence or PC+ copilot then you've not been paying attention. Scanning on your pc allows the government to side step encryption. None of your data is safe. It's game over for privacy.

GLaDOS isnt ai.

linus getting cancelled by the bots

10:48 I tried to reproduce that ChatGPT error you had with the numbers of r's in the word strawberry and found another interesting thing.
I did it with f's in the German word Dampfschifffahrt and it wasn't much better. It first said 3, then 2, then finally 4, then 3 again, and then back to 4.
But what's really interesting, the second 3 and 4 were in the same message without any interaction from me. At first, it wrote the word correctly but still gave the wrong answer. But when I asked for a deeper analysis of its "thinking" process, it also spelled out the word, only used 2 f's after the i, and then claimed that the word only has 3 f's. It then noticed mid-answer that it was again the answer I explained was wrong earlier, and the answer continued with the correct analysis and answer.
I guess that's like that because of the nature of LLMs where they always guess the next word based on the previous words. It concluded that the word has 3 f's and then based on the earlier words, words that already stated that 3 is the wrong answer, the next words should be words of a correction.

Tech bubble 2.0 is about to pop. Yay INTRO!

To nit-pick on an otherwise great video, you make it sound like Reinforcement Learning is something of a catch-all approach that can be slapped on to every ML method. It is not.

Generally, you have either supervised learning methods, where, as you say, a model is trained to fit to the training data distribution (text, images, sound, any type of information, really) and thus can predict or generate some information.
You then have unsupervised learning methods, where models are trained on unlabeled training data that has not been manually annotated.
Among the alternative learning methods to that you have Reinforcement Learning, where you have a closed environment and the objective is to predict the best policy (series of actions) that maximize your reward. An example would be the game GO, AlphaGO is an applied Reinforcement Learning model.

The premise and fundamental objective of Reinforcement Learning is distinctly different to other ML methods and it is used for different tasks.

so neuro-sama is an AGI then she's an AI vtuber made by vedal

5:20 Yes, neural networks are not really something new, and also old video games like Half-Life 2 already used AI. But as you already described earlier, AI does not equal machine learning and neural networks.
I'm pretty sure you mixed something up there. The Valve guy in that video is not talking about neural networks, he's just mentioning AI. We're at a point where that's changing, but the AI in video games we saw until now, didn't use any neural networks. They're all just simple rule-based AIs, probably with a lot of state machines.

Skip da ad 1:10 and 14:23

This has nothing to do with AI. It happened with the internet, crypto and now AI. People are missing the forest for the trees. Every new big paradigm change in tech has opportunists and scammers salivating over taking advantage of the opportunity. To think this means there are no core breakthroughs behind these is the most foolish thing one can do

The most disappointing part of this video is right at the beginning where linus is claiming the 'definition' has changed, when in reality, even A* pathfinding algorithm was always AI, and the kind of AI that movies referred to was always 'AGI'. Kind of disgusted by this representation tbh

Few people actually realize what an LLM is.

Predicting the next word is just a clever and brilliant idea allowing a bunch of neurons to get an idea of ​​the world and how it works, predicting the next word is just the tool that will allow the model to be “sculpted”.

This is what the human brain does, with more power the intelligence of LLMs will be indistinguishable from ours.

Just as translation, grammar, syntax are emerging properties of LLMs, reasoning will also be, and it will be more and more precise and fine as the power increases, until it surpasses us in this domain…

They don't spend billions for nothing, they have the trajectory.

Wouldnt call steeling date for art cuz all art is stolen

Please be careful on your personal information…

Linus casually dropping some truth pills at the end. I love it.

Thank you so much. I already knew that its not real AI, and more like machine learning. However, if you see it almost daily, you kinda begin to believe it. Thanks for affirming my knowledge xD

Prove me, people are not ANI

Machine learning is the process for creating an AI model… it's not what AI should be called, but the process to create it.

I want to buy AI screwdriver.
Hope it screws better than my ex.😂