You have more faith than I do. The current AI push in all these companies is not exactly a conservative, careful approach.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
How AI will fail
- Thread starter rinaldo00
- Start date
You have more faith than I do. The current AI push in all these companies is not exactly a conservative, careful approach.
which AAA game in 2025 or 2026 made use at all of new unproven AI of the LLM type (the dlss and others proven one sure) or any new unproven tech at all ? Some ai generated voice for AA games did happen, but that tech has been used and well proven outside of games for many years now.
Maybe some internal tool try out was less conservative, but there so many check and balance for the final product to catch anything bad with them, that it is not really risky.
Let just say it has been a while I heard the critics that modern AAA video games studios are not risk averse enough and too bold.
Had a run-in with AI, just plain google searching when it offered an unexpected unasked opinion. Sense enough to merit a reply and explain the problem better. Then it tells me it can read schematics. Turns out it can, and understands them very well. Most of it's suggestions for improvement were BS, but greatly improved from its first try. I didn't have time for further uploads with detailed explanation. It said we could continue later and would save drawings. Later, I returned to tell me does not save anything between sessions. Never again will I get sucked into that lying waste of time. After blowing my mind by thoroughly understanding uploaded drawings, I was inclined to suspend too much disbelief. AI is still inane.
Last edited:
If you're signed into your Google account when talking to it, it saves the sessions - I don't know about any uploaded images/files submitted during those sessions though - I realized this when asking it a question and it referenced a device of mine (and said "like your x device") - and I was like - "I haven't said crap about that (this time) how'd you know I have that?" - and it informed me from the time I asked it about that thing months ago (which I did), it saves sessions, yaddah yaddah.
- Joined
- Jul 13, 2006
- Messages
- 2,267
AAA games wouldn't have, yet, no. Anything coming out in 2025 or 2026 would have been in development for a long time before the current rush to include AI in everything.
- Joined
- May 18, 1997
- Messages
- 58,426
Triple A gaming companies are very much looking at implementing specially trained LLMs for NPCs. But no, nothing like that yet in the AAA space.
- Joined
- Jul 13, 2006
- Messages
- 2,267
stashinsome
Weaksauce
- Joined
- Jun 12, 2026
- Messages
- 106
not sure how i feel about the dead internet theory becoming fact so quickly.
- Joined
- May 18, 1997
- Messages
- 58,426
That guy is an idiot. Delusions of grandeur.
- Joined
- Dec 5, 2018
- Messages
- 5,099
He has no idea how AI works, where it's going, or anything else related to AI. His little demonstration of AI music was already embarrassing enough. He needs to not talk about AI at all on his channel.
- Joined
- Jul 13, 2006
- Messages
- 2,267
What is the 'too long, didn't watch' summary? Some videos have an AI summary, but I couldn't find it and I don't want to watch it.
"things are bad (memory/storage prices, certain groups sentiment towards AI) so I was right"
That is a very unfair summary. He may be an idiot and not understand AI well but he does spend most of the video discussing running local LLMs. He also expressed his frustration over prices like we all have, but he did not use that to prove he was right.
that quite the confirmation bias, the fact a lot of harddrive went into datacenter is a sign they are not being build and that if they are will not be used ? because AMD as a $4000 system that can do 100 TOPS (instead of paying $20 a month for 200 months).... same goes with people booo AI in speech that mean they will build their own local AI system.... instead of continuing of using it via datacenter without even knowing they are using it.
Those graduation crowds have the generation that poll the most against AI technology but they are the one that use it the most by far... which tend to be a bit like shale gaz back in the days, when something actually work great sentiment around it can get bulldozed.
Last edited:
- Joined
- Dec 5, 2018
- Messages
- 5,099
Yeah ... kind of hard to miss.I mean the very title of the video is "I Was Right About AI"
I will try again.I mean, the very title of the video is "I Was Right About AI"
You said this.
I said this.
Your post makes it sound as though he used prices as a reason for being right when that was a separate thought. Whether he is right or not the rest of the video talked about doing AI locally and since his premise is that local use is why AI will fail he was right.
In summary, his first video claimed that AI will fail (meaning AI companies) because people will do local AI. In the second video he showed Lisa Su and others pushing for local AI and therefore his original premise was right.
Except cloud AI still exists and they're still building datacenters with plans for more datacenters, not stopping the buildout, not plans for less
So he still isn't (as of yet, at least) and is just prematurely sucking himself off with that title and premise
https://news.constructconnect.com/j...ort-construction-remains-strong-as-costs-rise
That a really strange way to formulate things, if AI is popular at home.. AI did not fail at all, it is a giant success story, video game did not fail when home video game system replaced big shared machines in arcades.
If it is a shortcut to say that powered by datacenter AI inference will fail as a business model, since the first video of 3 months ago:
https://www.cnbc.com/2026/06/05/goo...million-a-month-for-xai-compute-capacity.html
https://x.ai/news/anthropic-compute-partnership
Is there much in the last 3 months that confirm one way or an other that running on large system ai inference will not be popular ? Does not seem in the video to prove that shift at all, there is even points made that contradict that view directly (like hardware being in low supply because of how much goes into large building dedicated to them)
Please stop quoting only parts of my posts without context.
I never claimed he was right, I only explained why he thinks he is right.
And I explained why his explanations you explained were still wrong - don't defend him if you don't want to be refuted in what you defend him on
I don't understand why you have issues with me. I am NOT defending his point. I even said this " Whether he is right or not". I just want people that did not watch his video to get an ACCURATE summary of his point so that they can make up their own minds about his video
Quote my entire sentence please. I said:
"In summary, his first video claimed that AI will fail (meaning AI companies) because people will do local AI. In the second video he showed Lisa Su and others pushing for local AI and therefore his original premise was right."
Read this part
"AI will fail (meaning AI companies)"
Strix Halo/mac mini/mac studio/nviida digits/spark and others already existed 3 month ago. DGX station were announced a while ago (we had the copilot+ pc years ago):
https://nvidianews.nvidia.com/news/nvidia-announces-dgx-spark-and-dgx-station-personal-ai-computers
That not a new development since the first video that confirm much.
Datacenter AI is being pushed has well and they grew in demand since.
I have listened to both video, I disagree that he was proven right in the last 3 months (i am not sure which part of my message let you to believe I did not understood you or him), look how much google and anthropic are paying for datacenter AI capacity in their deals they signed in that time frame.
I posted his videos to get a discussion going, which it did. I have no comment on his opinion, I am only the messenger. I thought you had the impression that I agreed with him based on how you quoted me. I see now that you are indeed only talking about him.
- Joined
- Aug 7, 2004
- Messages
- 19,734
I just liked how Rick asked Jimmy Chamberlin to play a bit of Jellybelly when he had him on for an interview, and Jimmy ends up just ripping the whole thing, lmao
View: https://youtu.be/PuZFyTvMPLk?si=aHH6HYEkb5f7c0Cy&t=254
View: https://youtu.be/PuZFyTvMPLk?si=aHH6HYEkb5f7c0Cy&t=254
- Joined
- Jun 7, 2007
- Messages
- 14,416
When you get in a groove you don't just stop half way.
- Joined
- Aug 3, 2004
- Messages
- 24,312
Ya, good luck with that..
~0.6% of the world GPD of 2030, yup at least, that already a bit old comments could be higher than that now.
Recommended digital ads alone market could be higher than that mark before 2030.
He must be talking about some narrow subset use of AI revenue that need to reach such a small number (and yet to do so), like API usage + subs + others in anthropic sphere type.
Recommended digital ads alone market could be higher than that mark before 2030.
He must be talking about some narrow subset use of AI revenue that need to reach such a small number (and yet to do so), like API usage + subs + others in anthropic sphere type.
Last edited:
- Joined
- Aug 3, 2004
- Messages
- 24,312
- Joined
- Aug 3, 2004
- Messages
- 24,312
He does not seem to think robotic-selfdriving and so on would be impacted by AI in 10 years and seem to be quite focused on text base LLM (which seem quite out of touch), but people underestimate how big is 1.1-1.5% of GDP is (that still a ~1.6/~3 trillion type of affair), that an old mid-2024 estimate.
nearly all the normies (and people online even in this forum included in that) do - that is what AI is to them they don't even know or care to know of the other uses/models/etc - it's 'just all chatbots' in their eyes
To be fair, that's what these companies are primarily pushing to the public so that's what the public knows. The other products are in the background and not being heavily advertised to most of the population because they're industry specific. If you work in medical imaging you know all about the AI use there but most of the public is clueless. If these people were actually spreading the advertising dollars around and showing everyone what some of the data centre load was serving then maybe it would help their image. But, they won't do that because they want to bump those token numbers as high as possible and the more normies they have making funny pictures of their pets the better for their token numbers. This is all to help train the AI and give a big token number to investors as to why they need those data centre dollars to be spent.
mullet
2[H]4U
- Joined
- Aug 19, 2004
- Messages
- 3,454
Can anyone explain the difference between the software running in a Tesla car vs a LLM?
the other side of that is you have models based on LLM (transformer) architecture/model/idea that don't do language/chat at all - genomics/protein folding/robotics/finance etc use customized LLM models (yes LLM models like ATM machine
) - and the development/furtherment of public facing 'normie LLMs'/chatbots/etc do help further these as well edit: it's the transformers really it's using from LLMs that are then customized, not the LLMs themselves
Last edited:
Same as I explain above - it has the transformer from an LLM but instead of chat it uses it to match to/decide things about physics/spatial awareness vs chat/language - so instead of in your head imagining blocks of words matching to one word/the next word/related word imagine it with things like tree/pedestrian/acceleration/etc - and it does these with images of the things like a tree or pedestrian - and it learns non-image things (like acceleration) from comparing multiple images of things (like a tree passing by) and figuring out 'acceleration' from the difference of the images and what must be happening in between them to make them look different
And just think of the transformer as the algo of matching/processing/predicting/etc
Edit: to add to that - so during training the model has no concept of 'acceleration' - it can only learn it by 'learning to run' and then 'running past a tree' to then see how that resulting 'mental head image' it acquires doing so matches the two different input images of a tree passing by - to then recognize the difference between those two photos is 'running by'/acceleration
and it learns to 'run' because at first in training it does nothing, then just sits up, then flops around a bunch of times, then gets up and falls over etc - and each successful model/run that is or closer to what the designer wants, is used for the next 'learn to get up and run' model/trial run - until through enough trial and error and 'survival of the most correct' - it learns 'running' - but also obviously with a Tesla you're giving it/the model the inputs/capabilities of a car not a human so it's 'driving' instead of 'running'
and it's not just one transformer/one algo/one matching process in total that's occurring when doing Tesla FSD - it's multiple going on simultaneously, all trained from multiple/different models - and then to coordinate/orchestrate all those different transformers/algos together in real-time is the FSD algo - which itself was modeled/trained on 'orchestrate and coordinate all these other transformers/algos together in real-time'
Double edit: and then to add even more and make it even more crazy - that initial image of 'a tree passing by' that it's fed - in order for the model to have a 3D arena to 'flop around and learn to run in' - it first has to actually design/create the 3D arena itself by 'randomly drawing/creating' the same as 'flopping around' until it accurately creates a 3D arena matching the input/tree photo that it can then learn to 'run past' until it finally matches and understands the tree photo - it's fed a bunch of descriptive/annotated and etc data attached to the inputs to help with this 'random drawing' phase (as well as the 'get up and run' phase)
Last edited: