AI Is Not A Bubble (Why Goldman Sachs Is Wrong)
Vložit
- čas přidán 29. 08. 2024
- Explore why AI is not in a bubble despite Goldman Sachs’ critical report. Understand the nuances of generative AI’s economic impact, infrastructure costs, and enterprise ROI. Learn about the broader perspectives and counterarguments to the skepticism surrounding AI’s transformative potential. Discover the evolving narrative around AI adoption, productivity gains, and the future of AI technology.
Learn how to use AI with the world's biggest library of fun and useful tutorials: besuper.ai/ Use code 'youtube' for 50% off your first month.
Concerned about being spied on? Tired of censored responses? AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit venice.ai/nlw and enter the discount code NLWDAILYBRIEF.
The AI Daily Brief helps you understand the most important news and discussions in AI.
Subscribe to the podcast version of The AI Daily Brief wherever you listen: pod.link/16806...
Subscribe to the newsletter: aidailybrief.b...
Join our Discord: bit.ly/aibreak...
Thanks for this good breakdown. I do agree that the finance bros are a bit salty about it, and they sound like people poo-pooing early computers. Still, there are a lot of great points in these articles and conversations. Well done.
Good to see you endorsing a video before I watch it. Feels almost like a stamp of approval.
They did the same with bitcoin. Now they are adopting, and endorsing it
I'm more interested in local open source AI than anything the mega corps have to offer for sale. Not to mention they had to neuter and censor it so much. I'm much more interested in running local uncesored inference and training. Planning my new home server build now for end of year.
The corporate tools can be useful, though. Local open-source is extremely important, but it is not necessarily the best choice in every use case.
One thing is for sure ... AI will reduce cost of writing software , documentation, learning new concept , searching for information. These cost reduction will have ripple effect on other industries and that is how it will impact. It may be 10 to 20 % extra growth over a decade but it is still a significant.
Before GPT-4 came out, it was all about the parameter count. Around the time of release they realized that they had reached diminishing returns on parameters. I suspect something similar will happen with scaling soon. After that, the ceiling of what LLMs can do might show up. If so, other techniques will be built on top of or augment LLMs. Or perhaps LLMs will be abandoned for some alternative approach.
It is already happening. New techniques in the field are neural symbolic self training, and evolutive learning.
SingularityNET is one of the leads in this front.
@@Firrodrigolosofia I've heard of them, pretty cool stuff.
I'm not surprised a business that fundamentally works in the domain of perceived value is scared of the change that is coming.😅
An argument founded on "we are investing long term and betting on AGI" is a weak argument. We have no reason to expect an LLM will produce AGI other than various over invested orgs promising it.
Until more compelling arguments with facts and supporting data come along I think the financial analysts win this round.
I don't really get what sort of data other than increase scale -> performance goes up you'd need. Until someone produces a larger model and it doesn't show serious improvements on broad ranges of tasks, the default to expect to me seems we do get AGI. I think the burden of proof falls on those claiming scaling laws, despite working for 15 orders of magnitude will randomly stop on the 16th order of magnitude.
What sort of data would you actually find compelling?
@@Alice_Fumo there are published papers explaining exactly how LLMs suffer from deminishing returns. You can literally go and ask ChatGPT to explain it, and cite some academic research. Furthermore the onus is on the data scientists to give evidence that AGI is a natural evolution of complex token prediction. I think the fact LLM fever is sobering up, and we are beginning to question if building a nuclear power plant to power chat bots makes sense should give you pause. What exactly do you think GPT5 will do for humanity? The cost of investment is billions, and billions, and unless it gives us the new wheel the bubble will most certainly pop. So far OpenAI is feeling a lot like vaporware. Their speech model is held back, sora is dull already, and GPT5 is nowhere. In fact, it seems like OpenAI is spending money figuring out how to make a more paired down GPT4 with GPT4o. I believe there is panic brewing as scale isn't delivering.
GPT-3 had a smaller carbon footprint to train than that needed to raise a college graduate from infancy, with an enormous leveraging in inference.
excellent analysis!
Absolutely true. As someone with significant lived experience using AI in my work (legal field)-this video is spot on. It is clearly made by someone with actual lived experience using and integrating AI to accelerate workflows in real businesses. The value/ROI is undeniable. Anyone questioning AI’s utility/ROI for enterprise is drawing a conclusion about AI’s utility based on adoption rates. The reality is that implementing new technology into existing workflows takes time and initiative. A little bit of legwork and initiative in this regard pays truly enormous benefits.
After covering the Adam Conover/Ed Zitron convo on my stream, I was so happy to have found this video. You did an excellent job here, and I appreciate very much.
Really solid episode. Appreciate the effort that goes into this, NLW!
Excellent work! Well done, so informative. 🙏👍
A very much appreciated and well done breakdown. I'm love this kind of deep dives, that don't just repeat the general talking points or contain just author's own opinions, but contain original analysis instead. Ofc both the talking points and opinions are important because this is YT, but it's rare to see someone really put work into their content like this, thanks :)
They are absolutely right. AI is a hugest bubble.
Did you watch the video? What is you comment even supposed to mean?
Lots of things to consider. On one hand the big tech will be alright, they do have more to lose if they sit it out. Apple did sit it out until June and the stock suffered, but they just added AI to iOS and suddenly it just blossomed. Apple spent comparatively little for AI and piggyback on OpenAI investment for heavy AI tasks. Sounds wise :).
In general, I don’t think that AI will add to the top line in a consequential way, it will be absorbed by the current product features due to commodification. But those who don’t integrate it, will lose out for sure.
So, the market is bubbly, not in an existential state.
Saying Gold Sackers will give away their money with straight intelligence to the retail market is about as useful as IBS.
The Gross Domestic Product is a quite questionable benchmark in the first place. It is also better to reasons about the economy, like in this video, than making calculations and models claiming these would predict the future or reality.
Very interesting video. Thanks.
Some AI stocks go bubble, but not AI as the industry. 🥰🥰🥰
Well done. That's the way to go!
He is absolutely right
Who is?
Thank you so much, your channel is so important to make people being correctly informed about AI World, i appreciate so much for your insights and your reviews. Specially for me, an AI engineer and student. One more time, thank you, Nathaniel, your job is very great and important!
AI has beaten kasparov (Chess)
AI has beaten Fan Hui (Go)
People seems to not understand what it really means to beat great great master in those things. There is exponential value in AI, it just needs refinement.
Anyone have any thoughts on TAO ?
Yeah, easily brainstorming, but human verified coding is awesome!
When the game gets "rigged" Goldman Sachs' is the "rigger" . You can't dismiss what they say even if it's not true.
There are a lot of parallel projects going on, so yeah, spending a 100x to basically invent the same thing a hundred times
Goldman analyst has not went to China where AI and robotics is driving the economy future as they face declining birthrate and resources constraint...
AI is a bubble, and it it not a Bubble. AI changing our lives to UBI soon, is probably real. Though it may be overlevereged rn
Gen AI is not a bubble, but it may pop their bubble.
A humanoid AI robot (hAIbot) helper will be invaluable for small farms, small businesses, homesteads and elder care. I have a lot of simple tasks around my farm, forest, butchershop and construction sites where hAIbots could be a big help. I need a bit more rugged than the Tesla Optimus. They need to be able to keep up with me: 6’ tall, lift 500 lbs for tipping and rolling hay bales, carry 300 lb at 4mpg walking speed, carry 100 lb at 8 mph walking speed, dexterous. IP67 weatherproof down to -50°F. They should be able to repair and service each other, At the often touted price by Musk of $15K I would take three for my farm. AI Robots, pitchforks, knives, chainsaws and lasers… what could possibly go wrong? 🤔
I think hAIbot is a cute term, but typically those are called androids. The contrast between a high-tech humanoid robot and doing a mundane non-modern task like rolling hay bales is kind of funny. I always appreciate seeing the real use-cases real people are imagining for this technology.
FUD like this lets me buy cheaper NVDA
I do not think any bubbe has ever felt it would even could be a bubble. if it did know admitting it might not be helpful in the current. admitting smoking kills at first was a bad idea.
YA see what I am sayin?
That said i really believe goldman is corrupt AF that aside AI is NOT a bubble.
Reasonable arguments fortunately remain reasonable, even when they are said by someone who has financial incentives to make those arguments. So to distinguish good from bad-faith argumentation, all we can do is look at the same data and see if we come to the same conclusions.
Just the arts will be threatened by ai
So we just need to wait a few years for genAI to give us more than text summaries, shitty insecure application code and generic imagery based on plagiarized art? And then a trillion dollar ROI will emerge?
The talking heads are trying to drive down stock prices in the months before ChatGPT 5 comes out.
You don’t understand what ROI means. It either helps the bottom line or it doesn’t. No one cares if it helps employees maintain the same output with less work.
I disagree, same output with less work means more bandwidth to quality check said output, which reduces rework or defects, which is a direct impact to ROI
@@ryanrobinson9732 You assume people will reinvest the time they saved into other tasks. Not my experience. And even if they did, that would translate into productivity improvement, which is still different from increased bottom line. Your spend is the same, unless you reduce your workforce proportionately. And it’s doubtful your sales will increase because the org is now using AI while producing the goods.
it is just a pump and dump scheme, do not buy in AI unless you fully understand all the tech specs.
Cope. Harder
Booh down with technology 😂
W🎉 w