There was a tech bubble, period. Still is. Lots of tech with inflated value, driving massive hiring that would never result in enough value to justify the jobs.
I know multiple companies hiring for teams with dead or dying projects. They hire anyway because if they don't they'd have to explain to the higher ups how the projects those higher ups green lit are useless. They need more people to make it seem like the stupid shit they're doing will generate revenue, when they know it never will. Keep the gravy train going as long as possible.
Most of it was/is around other bubbles, like the crypto bubble, AI bubble, B2B bubble. Crazy ad spends have been making it worse. The pandemic inflated some sectors and deflated others, but mostly tech was fine while other industries floundered.
I'm not convinced that we are in an AI bubble. In June 2007, Apple released the iPhone. Its stock price was $4.45 adjusted at that time. A year later, it was $6.50. 15 years later, it's now $153.
The iPhone is a real product, though. AI is more like an automated fortune teller. Just because your fortune sounds really convincing doesn't mean you should use it to buy lottery tickets.
I think you just haven't seen the potentially revolutionary applications.
My side project is developing a tool to help me learn Chinese. Last night I asked gpt-3.5-turbo to:
1. Take the existing text which had Arabic numerals in it (e.g., "2006年" for "the year 2006" and "30万" for "300,000") and replace them with Chinese characters as they would be read (i.e., "二零零六年" for the year, "三十万" for the number)
2. Segment the text with some specific parameters (i.e., segment "二零零六" as four separate words, rather than one big word)
3. Convert the whole thing to pinyin, with specific parameters (don't group multi-character words and don't capitalize proper nouns).
I got the output I wanted with 5-10 minutes of playing around with it; I could theoretically use those prompts to process my entire library of study content. Do you have any idea how long it would take me to code something like that up?
> Just because your fortune sounds really convincing doesn't mean you should use it to buy lottery tickets.
You shouldn't buy lottery tickets anyway (and ChatGPT will tell you so, if you ask it). I have difficulty to understand where this negativity (not speaking of other downsides) regarding AI comes from. Yes, it talks BS sometimes, but so do people all the time. And, no, it's not the solution for all our problems, but for some things it is damn useful.
I think there’s negativity because it gets hyped too much.
We were supposed to have self driving cars all over the roads 2 years ago. We still don’t have anything close but now people are all “How is the way AI works any different from the way the brain works. Looks the same in my non-expert opinion!”.
I think there’s negativity because it gets hyped too much.
You only can recognize the true revolutions in hindsight. I remember well the people who said the Internet was overhyped and it's just a novelty that will go away. In hindsight it was not hyped enough. My personal opinion is that what we see in AI is more like the early Internet than the hype about self-driving cars.
What I see more of an issue is that some people expect AI to solve all their problems and when they are disappointed they overlook that it solves some problems better than anything we had before.
This is true. I think it was hard to reason about how important the internet would be because it was expensive to get on and it needed Google to make discovery exponentially better. It took a long time for access to become cheap enough to be almost ubiquitous in a developed country.
ML, AI, whatever, is clearly usable now. But it has limits. And I think it's clear that it's not fully baked relative to the promises. We certainly don't have a Killer App yet.
What we have isn't good enough, and I say that at someone who in 2009 expected cars that didn't have or need steering wheels to be purchasable by normal people in 2019[0], but the tech does exist and can drive you around.
(And derive you around, as autocorrupt decided to inject a suitably ironic twist that I almost didn't spot).
> but now people are all “How is the way AI works any different from the way the brain works. Looks the same in my non-expert opinion!”.
In general: yes, totally. (Though I've only heard that specifically being used for "why stable diffusion is no worse than art school WRT copyright issues").
Always the way, though: "Consciousness is mysterious, quantum mechanics is mysterious, therefore they are the same thing".
> You shouldn't buy lottery tickets anyway (and ChatGPT will tell you so, if you ask it).
I was curious so asked..
ChatGPT:
As an AI language model, I cannot endorse or encourage gambling activities, including playing the lottery. However, I can provide some information about the lottery and the best strategies that people use to increase their chances of winning:
- Play regularly: The more you play, the better your chances of winning. But remember, playing the lottery can be addictive, so it is important to set a budget and stick to it.
- Play less popular games: Less popular lottery games tend to have smaller jackpots but better odds of winning.
- Join a lottery pool: Joining a group of people who pool their money together to buy lottery tickets can increase your chances of winning.
- Choose your numbers carefully: Some people believe in choosing numbers that are meaningful to them, such as birthdays or anniversaries. Others prefer to choose random numbers. It's important to note that the odds of winning are the same no matter what numbers you choose.
- Consider the odds: It's important to understand the odds of winning the lottery. The odds vary depending on the game, but in general, the odds of winning the jackpot are very low.
It's important to remember that playing the lottery is a form of gambling, and there are no guarantees of winning. It's important to play responsibly and never spend more than you can afford to lose.
I also asked it before I wrote my comment and the answer I got was much more diplomatic than yours. I used the brand new GPT-4 and here is the answer I got:
As an Al language model, I don't have personal opinions, but I can provide some information to help you make a decision.
Buying a lottery ticket is a personal choice, and it depends on your reasons for wanting to purchase one. If you're buying a lottery ticket for fun or entertainment, and you can afford to lose the money you spend on it, then it might be an enjoyable experience for you. Keep in mind that the odds of winning a large lottery prize are extremely low, so it's important to have realistic expectations and not rely on the lottery as a means to solve financial problems or achieve long-term financial goals.
If you are considering buying a lottery ticket as an investment or as a primary means to improve your financial situation, it's essential to understand that lotteries are not a reliable or sensible investment strategy. The odds of winning are incredibly slim, and you are much more likely to lose money than gain any significant returns.
In summary, if you want to buy a lottery ticket for fun and can afford the expense without any negative impact on your finances, go ahead. However, if you're relying on the lottery to improve your financial situation, it's important to explore more reliable and effective methods for achieving your goals, such as saving, investing, and financial planning.
I thought it fairly obvious that the poster you're responding to was making an analogy about using GPT/LLM models for decision-making, not actually endorsing playing the lottery.
And they're correct. Don't use generative models for extractive tasks. That's like using the handle of a screwdriver as a hammer. Sure, it might work for a while, up until there's a screwdriver in your eye.
Good point, and I wholly agree. It's not only that GPT/LLM models are not the solution to everything, as some like to pretend - some applications are even potentially dangerous.
You're saying this as if people weren't saying worse things about the iPhone than you're saying about AI. Steve ballmer said it was an expensive piece of junk without a keyboard that nobody would want. You just have to google iPhone and iPad critics from that time period to see that you're having your own Ballmer moment. If you're right or not only time will tell.
I use ChatGPT, GitHub Copilot, Midjourney, Fireflies, Cleanup.Pictures, and a host of other AI-based tools in my daily workflow. I pay money for much of that. Anyone saying that these tools aren't products are, to be frank, deluded.
For some reason, people are equating inaccuracy with redundancy. I can't fathom why people do so.
I have a copy-writer, editor, artist, co-coder, photo editor, reporter, all at my finger tips. I need to give them advice on how to iterate their designs, and I know not to ask them for explicitly "correct" information, but they save me hours of my life, and hundreds in hiring costs. I can get templates, corrections, suggestions, tips, etcetera, in seconds. It's incredible.
Most of my non-programmer white-collar friends use ChatGPT daily for their work. They are aware that it hallucinates stuff, so they don't ask it factual questions. Instead, they use it to transform existing data. It's wonderful at that task and saves them a lot of hours.
If you want to use that logic, why aren't we talking about banana farming? Also sold, also a product.
The biggest difference is one is an ongoing service of questionable utility to most people in it's current form. iPhone is pretty instant. I have an ipod. I have a cellphone. Now I have one device. Chatgpt can't really help me on the day to day right now. I'm sure as hell not paying for it. In 2008 if I could have paid for it, the iPhone solved a real thing I was facing and I would have purchased it.
You might not pay for GPT-4 but my employer would gladly fork over many thousands for each of our tens of thousands of employees if we could train it on our own dataset and run it in-house.
"In its current form" is also a weasel word. GPT-4 will replace hundreds of thousands of jobs over the next few years. It's already disrupted search, it'll soon disrupt most fields.
I see iphone being used by everyone. Trains, buses, even beggars have iPhones. Do you honestly think AI will have that level of penetration? As in not as piggy backed on another platform? That’s the only way to extract a lot of value.
There is a few nice improvements here and there like iOS handwriting recognition (similar function existed before AI) and clipping picture, but the products already existed for decades and AI only bring a few incremental novelty.
On the other hand, algorithmic feeds ands ads have made social networks way worse than they were a decade ago, and spamming is now on another level thanks to GPT.
So for now positive contributions of AI have been tiny and balanced by the negative ones.
• A photos app with an AI that identifies image subjects and allows search by content
• Computational photography
• 17 language translation built-in, more if they get Google Translate
• OCR when you tap on an image containing text
• Voice-to-text transcription in the default keyboard
• Fall detection and step counting powered by machine learning
I think AI is likely to be like "3D" or "multimedia" in the 90s: it will be a sales buzzword until it's everywhere and therefore no longer a differentiator.
3D acceleration itself is still a big market, it's just that when everyone had a GPU (unlike my Performa 5200), the marketing shifted to which specific brand — NVIDIA, AMD, or whatever.
I'd say a fundamental expansion in how much people use computers is what made smartphones a big shift. Maybe conversational generative AI will have the same effects, though I'm not too sure about that, people already use computers a lot.
I know multiple companies hiring for teams with dead or dying projects. They hire anyway because if they don't they'd have to explain to the higher ups how the projects those higher ups green lit are useless. They need more people to make it seem like the stupid shit they're doing will generate revenue, when they know it never will. Keep the gravy train going as long as possible.
Most of it was/is around other bubbles, like the crypto bubble, AI bubble, B2B bubble. Crazy ad spends have been making it worse. The pandemic inflated some sectors and deflated others, but mostly tech was fine while other industries floundered.