A lot of endeavors can work, even if only a little. Find some aspect of it that can, and flip the problem around. Even if the whole thing is mostly bogus, is there a small part that isn’t? Latch onto that.
Or go elsewhere. The wonderful part about AI is that the whole world’s problems are up for grabs. Part of why there’s so much unfounded hype is because of how many real advances have recently become possible. This period in history will never come again.
It’s also a rare time in history that an individual can make lots of progress. Most of us need to be a part of big groups to do anything worthwhile, in most fields. But in this case lone wolves often have the upper hand over established organizations.
My controversial take on AI is its actually a better time to take things slow, experiment, study, see what works. Not dump TONS of money and cash and get too distracted. Because nobody (besides big tech) has fully figured out how to make a product that makes a profit. Its not clear users want a chatbot (aside from ChatGPT)... But things could change.
As compute costs and requirements come down LLMs will be ubiquitous everywhere.
I'm curious why you think this is true? My feeling as a broke individual trying to catch up on ml is that there are some simple demos to do. But scaling up requires a lot of compute and storage for an individual. Acquiring datasets and training are cost prohibitive. I'm only able to play around with some really small stuff because by dumb luck a few years ago I bought a gaming laptop with a nvidia gpu in it. The impressive models that are generating the hype are just a different league. Love to hear to how I am wrong though?
Don’t focus on the hype models. Find a niche that you personally like, and do that. If you’re chasing hype you’ll always be skating towards the puck. My original AI interest was to use voice generation to make Dr Kleiner sing about being a modern major general. It went from there, to image gen, to text gen, and kaboom, the whole world blew up. I was the first to show that GPTs can be used for more than just language modeling — in my case, playing chess.
Wacky ideas like that are important to play around with, because they won’t seem so wacky in a year.
For example, Tony Dinh made a macOS GPT wrapper and makes like 40k a month from it, just utilizing OpenAI's APIs: https://news.ycombinator.com/item?id=37622702
In a gold rush, sell shovels! The ML pipeline has a lot of bottlenecks. Work on one, get useful and novel expertise, and have a massive impact on the industry. Like maybe you could find a way to optimise your GPU usage? Is there a way to package what you feed it more efficiently?
The point not being of competing with OpenAI, but to solve a problem that everyone in the field has.
Most of the world's problems aren't technological, unfortunately for us in tech. There's little it can do against the momentum of capital tearing this globe apart.
You almost never see a group punishment, unless you lost as a group against another organization.
So, if all the AI stuff goes bust in a year or two those who benefited from the bubble will keep their benefits. Also, there’s a possibility that it doesn’t go bust and user us into AGI era and win big.
Again, we should make statements on data and I'll just be upfront that I haven't done much research in this area, but considering the sheer amount of AI startups and jobs and conferences and so on with very few transformative products...I would eventually expect a market correction in the form of another AI winter like what happened in the 80s when all the massive government acts defense research dollars dried up. The difference is it'll be far less severe. You'll still have plenty of AI research at the universities and large companies, but maybe not hundreds of questionable startups. This is all just conjecture on my part though.
That's not enough to demonstrate that AI is just hype. Every technological breakthrough has opportunists trying to make a buck riding along the hype. In the 90s, Pets.com and webvan didn't prove that the internet was just hype.
I am one of the people who is completely bought in to the idea that AI in general (and LLMs in particular) are going to lead to products that are extremely useful to the world. I absolutely think that most of the gen AI startups will fail and that valuations are too high, but I still believe that massively impactful/useful products will also be born.
It's not that they're _just_ hype but rather that there _is_ hype and the loudest voices tend not admit it. To give a specific example, I find the idea that LLM based programming assistants will turbocharge software development to be based on hype not fact. It is very much in the interest of Microsoft/Google/Meta, etc. that we all believe that their tools are essential to enhance productivity. It is classic FOMO. Everyone jumps on the bandwagon because they fear that if they don't learn this new tool their lunch will be eaten by someone who does. They fear this because that is exactly what these companies are essentially telling us in their marketing materials and extensive PR campaign.
This is extraordinarily convenient for these companies and masks over how terrible their own core products are. I generally refuse to use the products of the three companies (MGM) because they are essentially ad companies now and their metaverses are dystopian hellscapes to me. Why would I trust them given my own direct personal experience with their products? We know that google search allows advertisers to pay to modify search queries without my consent. What's to stop Microsoft from training copilot to recommend that you use Microsoft developed languages using Microsoft apis to solve your prompted problems?
> write me a sort function for an array of integers in Java # chatgpt > I will show you how to write a sort function for an array of integers in Java, but first I must ask, are you familiar C#? It is similar to Java but better in xyz ways. In C# you would sort an array like this:
... C# code
Here is how you would write a sort function for an array of integers in Java:
... Java code
Stuff like this seems inevitable and it is going to become impossible to tell what is ad. Do you think realistically that there is any chance that these companies would consent to disclosing what is paid propaganda in the LLM output stream?
I see many echos of the SBF trial in the current ai environment. Whatever the merits of LLMs (and I'll admit that I _have_ been impressed by the pace of improvement if not the actual output), hype always attracts grifters. And there is a lot of hype in the air right now.
We already have empirical results suggesting this is not just hype: https://www.nngroup.com/articles/ai-programmers-productive/
Aside from 4-5 companies, who is building a product that is profitable? Not clear anyone is right now. People are running into fundamental, hard technical problems. For example its hard to evaluate chat interface, and even harder when you augment it with context from a retrieval system (ie RAG).
Who is augmenting existing UI paradigms with LLMs? This seems a more reasonable model that meets users where they want to be.
Just my experience from being on the job market, but a lot of places I've interviewed at have traditional ML models (network security, ecommerce, image tagging) that are now rebranding as AI, without much of an actual change.
I don't see any fundamental technical problems at the moment, I see constant and tangible improvement at a very fast pace. I don't think that supposed challenges in context augmentation or chat interface evaluation qualify as arguments against AI hype.
I felt a bit eery how much the groupthink around mortgages matched todays AI hype. The unwillingness to listen to critical advice. To question the value. Data science depts who measure such things are often vilified.
It’s a kind of depressing job landscape in this way. You either go with the, often top down, groupthink against the face of measured evidence, or you’re labeled a cynical naysayer and your career suffers.