I can’t catch streams but I hope someone mentions the fact that NVDIA is losing so many engineers rn cuz they’re all basically retiring from all their stocks rn and it’s funny af
The 35 year old junior engineer that retires off of millions instead of staying and eventually becoming instrumental senior staff is going to be a big issue in a decade or so. Good thing the market only gives a fuck about what will happen next quarter.
CNBC anchor on air said "You'll be able to tell AI to show you emails from 'Devin' sent between 2021 and 2022..." That's just search. Every email platform/client does this. I know TV hosts aren't the people spending on AI, but if that's not a sign this is going to quickly become nothing I don't know what is.
The AI buzz mimics a lot of what was going on with the NFT bullshit. People were claiming all these useful benefits and features and they were being put into everything... then AI happened. Soon something else will take AI's place.
The problem is that AI is the buzzword to end all buzzwords right now. Most people don't understand the technology at all and are bullish, despite there being no clear pathways to wide scale profitability in the short term, precisely because they don't understand it. It's this vague nebulous term that everyone believes will revolutionize this that and the other, but nobody actually knows how.
Well, no, it's a bit more than that. Yes, you could ask for that. "Please show me all emails from Devin between 2021 and 2022" But this isn't really any more useful than existing email client tools. It gets a bit more interesting when you say "Please show me all emails from Devin talking about my plans for my deck" And where this gets a bit more interesting than it sounds is that this isn't as simple as a CTRL+F; it's possible an AI model could find emails talking about your deck without using the specific keyword (ie: "I just ordered some wood the other day for **it**"), for sake of example But this isn't even the final form. Where this gets really crazy is the agentic workflows. You can ask an LLM to return JSON formatted text, or predefined functions, and then on your receiving end, when you get that response, you can parse it and execute a function based on that response. So, for example, if you ask it "What are the main tasks I have left to do this week" The LLM could respond "check_email(2024, July, 22, 27)" To run a function to check all emails between the specified period, feed their output into the LLM's context, and then summarize them, while looking to you like it just said "According to your latest emails, your remaining tasks are... " ...But what happens if you have, for instance, a piece of software engineering you need to get done? Imagine saying "Please produce a progressive web app capable of hosting chats between several clients", and the model returns "run_vision_OCR_instance" "run_engineer_instance" "run_code_test_writer_instance" And so on. You could essentially have various instances of the same LLM focused on executing specific tasks, breaking up the task into appropriately sized chunks and checking their work along the way, and doing all of this while you're focusing on other tasks, only checking in with them every now and then, while the model essentially prompt engineers itself. In other words, it's a lot less like a chatbot, and more like a junior coworker with relatively good knowledge of the specific syntax of a variety of programming languages. But, if current models are like a junior coworker, what do models and tools like in three years? Something to keep in mind, is that this isn't limited to software engineering, or even "data" jobs that live in the digital world. You could imagine similar technology being paired with future work from someone like Jim Fan, to produce a reinforcement learning program that let a robot operate in the real world, with moment to moment directives being issued by an LLM agent as described above. "Check the oil levels in all the machines" "Sort out all of these packages" "Mark off this area of the quarry for mining" "Check the structural integrity of this weak are in the tunnel" These aren't sci-fi. They're not ready yet, but they're possible with today's technology, we mostly just haven't finished the optimized software for these use cases to operate "magically" and unintrusively. I totally get that there are a lot of gimicky uses of this technology floating around right now, but it's about to get insane in a very short period of time now that the software industry has frameworks that are almost the right fit for making this stuff happen, and agents are going to be able to be given goals, with the ability to operate somewhat independently from their handlers for a period of time without being micromanaged. If you haven't used this type of technology yet and it sounds really abstract and useless, I absolutely understand, because it's kind of weird to be confronted with it suddenly if you didn't follow it being developed, but we're only a few small jumps in error correction, memory, ICL, and possibly scale, from these technologies being disruptive...And I don't think it's going to stop at just being "disruptive" or just being applied in "a few industries". I think in 7 years everything will just be completely different.
Hey Big A, saw your question about where AI actually provides value. I work for a legal workflow solutions company (SaaS for law firms, legal counsel in corporations etc). AI is actually huge in the legal software space. The reason for this is because so much of legal work revolves around massive corpuses of text (generative AIs bread and butter). It can provide court summaries, quickly find related cases, draft contracts, edit contracts to be more/less favorable for one party (particularly useful in M&A law) and much more. I think AI is mostly a bubble, but I genuinely think it will make a massive splash for legal professionals. Interesting to see how headcount looks at these massive law firms a few years down the line
It also causes issues for lawyers who use it as it often hallucinates cases and summaries. This has been in the news multiple times already. Remember the AI is programmed to give you the desired result, regardless of if it exists or not.
I noticed the orange juice thing up here in canada a few months ago. They renamed it to "orange breakfast" and they have big pictures of oranges on the box with tiny slices of apple and grapes behind it lol
My school had a hackathon where all the hard theory based projects lost two some shit openai api project that took some pdf and spit out a summary. Not a single person was happy.
That's wild considering all the leg work is being done by openai. Like you can already just do this using ChatGPT, so what was the point of the project?
@@Jaydoff All i know is i spent 36 hours trying to prove the space complexity of a board game to loose to someone who used the open ai library and a react website.
Jesus that 5y Nvdia growth chart. I imagine there was a PC gamer 5/6 years ago going, "These new-fangled RTX GPUs are hella sweet! Let me put a thousand bucks into Nvidia just cause they're so cool!"
@@th3summoner first, our market cap is like 118 Billys. Nothing special anymore. I don’t think you understand how dire our situation is economically. Aside from our chief executive getting a 45.2 percent raise, we have bled a good majority of engineering and manufacturing employees in the last two years. All the new salary employees cost more and know less. Just in my business unit alone we have 6 employees under the age of 26. It’s an engineering org. This company is fucked, and most of us are aware. Furthermore, July 17 is when they vote to sanction a strike or not, and I don’t see how they vote no. Even our best defense programs are being shredded by the FAA and DCMA.
Saying this as a software engineer, video streaming is a technically difficult product. To get it right you need smart engineers. Netflix pays top dollar and hires top engineers. No other streaming platform is willing to pay engineers enough to get the smart people that are needed to solve the technically difficult problem that widescale, profitable video streaming is.
This. Netflix has been at the forefront of software engineering and operations for the better part of the last decade. They're still a soulless company that will gladly fire you if there's any sign you're under performing by some arbitrary metric, but their engineering culture is still one of the best.
That’s not cowardice, that’s just reason. You made a logical financial decision and there’s no shame in that. Don’t linger on the past man, there’s always opportunity in the future 👍
The market is currently fueled by fomo and greed. Those are miserable emotions to run on as a human. I think you'll find yourself much happier if you don't let the fomo and greed run you as well. Leave the casino to the gamblers
AI has a lot of specific use case scenarios. One good example is when a company has a large database of internal documentation, you can use RAG(Retrieval Augmented Generation) to ask questions of a chatbot to get answers from your own documents without having to search for 3 hours. The problem is RAG uses vector databases which require a whole new skillset that humans have to learn to implement, and we are doing that, it's just going to take a little time before end users start to see real tangible results that improve their lives.
I’ll add to the orange juice thing, down here in Flawda, a lot of the fields soured many years ago from severe cold snaps. Most growers did not replant orange trees. They instead kept the property and in the past couple years a lot of them have been sold, demo’d and developed. And the remaining legitimate orange groves are also being developed. In my home town, when I was a kid, everywhere was orange groves. You’d smell the blossoms driving through town. All the roads were single lane. Now everything is 3 lanes, there’s toll roads butted up to the outskirts of town, parking lots, strip malls and tract home subdivisions. It’s really sad. I’m in Lake County, largely considered Orlando market.
The "habitable" zone for oranges has shifted up towards Georgia due to global warming, the same people who close down their farms because of this reality fucking go out and deny it. its insanity.
I started believing in Ai after I used it to practice history taking ( I am a Medical student) I told it to come up with real life cases and then let me ask questions and reach a diagnosis, it helped alot and I really got to see the value of ai
One of the cool things that i saw come out of ai is an app that tell you how much carbs there's in your food by taking a picture to help diabetic people with their insulin doses
Big A you should talk about the energy implications of AI usage because I saw some stuff online that suggests it’s very energy intensive. Just wondering about what this could mean in the long run.
I think ai offers a significant timesaver for certain jobs and workflows... but its taking a machine that was going 60mph and making it go like 75 mph... in a world where we are already so productive that there are less jobs than people.
I been thinking this for a while, it doesn't make much value. I think it will have some value but for now it is vasty VASTY overestimated. I still think the stock gonna stay high from legacy value
4:13 "Large Supermarket Chain in Australia" lmao there are like two options there green and red (maybe Aldi, but they wouldn't waste money on that shit). Feels like something woolies would dump more money on to play "catch up" with Coles and their so-called decade of AI experience. Like the 400 million dollar rebrand of countdown in NZ. Unbelievably tone deaf corporation.
Def right on the "calling things AI". So many websites' chat feature are like "I'm an AI, you can ask me whatever" but it's not smart, you can teill it's still basic scripting (ie you mention x keyword, they respond with x default info).
I don't think AI is a bubble though, the technology is real and once AI gets smarter it has potential to really change the world. Thinking about how good AI is right now is just missing the majority of the picture. I know that sounds like a bubble ("think of the future!") but the science here (fitting predictive models to AI performance on datasets, for example) really supports explosive AI growth continuing. AI is *not* moving slower than the scientists predicted. It's just the random hype people, CEOs, and lay stock buyers who are overpromising.
could say the same thing about the internet in 2001 before the bubble pop - it's about timing. we all know it will be valuable eventually, the question is will it be AS valuable as promised AS soon as promised.
@@AtriocClips There are a bunch of overpromising companies (Apple, Google, Microsoft, etc.) who falsely claim their AI will create a lot of profit soon, and I wouldn't invest in those. But what you don't realize is that even if they don't make profit, they will still continue to make bigger and bigger training runs, especially the actual leading labs like OpenAI and Anthropic. This means Nvidia and compute providers will make profit even if the AI companies aren't making profit yet. For context I'm working in the field and regardless of what lay investors think, the actual people working on it think AI will start to explode around the 5 year timeline from now. (As in get qualitatively good enough to actually dramatically shift our society). Basically, the lay person's lack of understanding of how rapidly AI is advancing (only a very narrow slice of AI experts actually understand this) cancels out with the overpromising of normal bubbles. TLDR it doesn't actually matter if AI makes money at all in the short term, there will still be exponentially increasing compute demands.
Bought nvda late, it went up 15% Big A said he sold. I thought about it, it's an unsafe and speculative long term and sold. It's starting to fall and AI isn't paying off yet.
could you imagine a series of 3-5 hotdog eating competitions with several rounds of hot dogs and each season has a meta based on type I.e 30% vegan, 20%????, 50%Beef
I'm confused, I see a lot of clips where he talks about Marketing Monday, but I don't see any marketing mondays. Not criticizing, just confused wondering if I've missed some
I know this is probably cringe, but I wish you hosted a podcast or some form of longer media about this kind of stuff. I'd love to see you bounce these ideas around with co-hosts and/or guests.
Hopefully Apple can spin this retaking of the #2 spot into hiring more lawyers that will get bored and decide to get apple juice outlawed for trademark infringement for the lulz. That seems like the most realistic scenario for preventing all liquids from turning into apple juice at this point.
So for the Joey Chestnut thing, I believe the conversation went something like this Nathans: Hey Joey, so you don't have to directly promote Nathans, but you can't promote a competitor either Joey: No Nathans: What if we don't even promote ourselves so no brand is being recognized beyond just the event name? Joey: No Nathans: Joey, we can't allow you to promote them on our stage Joey: Deuces brah I can't remember my source so it's basically "just trust me bro" but from what I heard Nathans was trying to bend over backwards to not lose him while still saving face.
MKBHD made a great video about AI as a feature vs. AI as a product. Marques argues that AI as a seamlessly integrated feature is far more beneficial than AI as a product. For example, the Rabbit R1 and Humane AI Pin cannot stand alone as products because AI in isolation is not a beneficial product. But Apple Intelligence using AI to help you find photos and emails is extremely beneficial
no its just trader psychology. Think about when toilet paper got expensive during covid. It's just that everyone joined in on doing the same thing because of fear and then the prices made no sense for the same product.
Real question: aside from just not being as good of a product, why aren't AMD able to capitalise on this AI bubble. Why is NVIDIA the biggest company in the world but nobody even bothers taking about AMD. Are their GPUs genuinely just that fucking bad?
It's not because of their GPUs, the issue is with data throughput. For gaming AMD and NVIDIA are tied. But NIVIDIA had CUDA. That's the difference maker for AI.
Nvidia spent a long time behind the scenes developing ai specific hardware and got a massive jump on amd. In ai applications, they really are just that much better. Also Nvidias main product is gpus where amd also makes cpus which they have been very competitive with over the years so I think it's just a difference in priorities
Also, Nvidia has worked really closely with universities to teach new graduates Nvidia specific technologies for AI. This means that when graduates leave college they only want to use Nvidia AI platforms, which definitely plays a part in increasing demand in the industry
You say Nvidia needs to find a use for AI? How about total world domination "Terminator 2 style". They rebrand as a weapons company using AI for humanoid soldier robots and missile systems... That eventually take over the world. Nvidia will take blood money.
things are worth what people think they are basically. if people think randomly that nvidia is worth double then the chart will go up lol. Same for dollars actually.
I think you’re not on the right side of this. So many sectors will be obsolete when a semi competent business owner can replace an employee with a prompt to an ai
I think you might have a misunderstanding of what AI is. If its a code making decisions, even if its hard coded to always make the same decision, it is Artificial Intelligence. Artificial being the keyword. While machine learning / deep learning / and generative AI are just subsets of AI which are the "current" innovations in AI. So when someone does a spreadsheet that automates work it would still technically be AI, not the one they are trying to hard sell you on now.
Unfortunately Applied Natural Language Processing is not quite as marketable as the simple, memorable, and recognizable AI but sadly I worry people will begin to colloquially think that that's the only thing that can be called that... Language is important but marketing isn't always from the perspective of the benefit of all unrelated entities. You market to sell a product so naturally not all of that will necessarily be good for others since its not an inherent goal of it. Also probably a bunch of people doing pitches to companies / execs / etc. are just using AI because they don't understand it either and maybe that's part of the problem.
He is not misunderstanding. He is talking about the current marketing hype around Generative AI, LLMs. Big tech companies have branded it as such so he is using their language.
This "AI" bubble will completely pop in a short while because it's the exact same grift as crypto and nfts. The energy consumption for the computers running that garbage generative ai is absurd and should be a crime.
I put 10% of my wealth in NVDA... I don't think many people have heard of NVDA. Ask people around you, especially women, and they'll tell you they heard of Apple, Microsoft, Google, etc. but not NVDA. We're in a bubble only when most people have heard of the company
@@Broken_Mesh #1 Semiconductor manufacturing plant that is backlogged 10 years by NVDA, Apple, AMD, Intel, Meta and Google, still a good buy and I have a 5% holding that is my best return so far under apple