Generative AI is unpredictable.
Generative AI isn't that creative.
Generative AI doesn't reward actual creators.
There's legal questions about it.
We have public semantic search for years and it's good.
Imagine semantic search for your file system, ticketing system, code repo, document store, database, email.
That's where the time saver is.
A search system is always an ensemble of many parts that make up a quality experience. Generative AI will undoubtedly be a part of that from now on.
Look up the retriever/reader pattern to learn more about this. It’s a really interesting field and I’m really glad to see more people getting involved with search tech with the recent AI boom.
Edit: I should also shameless plug the book AI-Powered-Search on which I'm a contributing author. Just looking at the table of contents you can see how deep this subject goes, and how much more there is to learn! https://www.manning.com/books/ai-powered-search?a_aid=1&a_bi...
Trey is currently writing more on the Generative AI stuff, and the three of us are giving a panel discussion at the upcoming Haystack search relevance conference https://haystackconf.com/2023/
In person, when there is no algorithm, social norms ARE the algorithm. Maybe that'll work here too, maybe not. Good luck.
1. Carve out a career and build some useful tech with it
2. Get VCs to give you money to build a company that gets acquired
3. Build a genuinely profitable private enterprise with it.
Etc.
Multiple possibilities for (professional) success here which doesn't involve generative AI being bigger than semantic search.
But that's probably just my elder millennial desire for stability that's beyond my reach.
Many were/are rightly concerned with proof of work mining, now ontop of all that there is an even bigger incentive on a global scale to have a million high powered GPUs running all the time. And then they tell us these things will completely replace not only index-based search, but education, customer service, all document processing, etc.
Even if these things can underpin a viable business, its going to still push the needle that much further into unviable planet. Which, I dont know, doesn't feel like a particularly sophisticated future to me.
Semantic might maybe could be the way to divide a solution up among a set of competing models. Right now, the real pain in the butt of using models for solutions (for me, anyway, aside from the unpredictability and explainability and sheer size and corpus) is that a wrong solution requires another trip back to the model (or, I guess, mastering the art of "prompt engineering", but I haven't gotten consistent guidance on how that works). Chopping up the parts of the solution, then using smaller models to compete for the chunks, then assembling the chunks and smaller models compete for that - this could be where we end up, and semantics might be one possible way to chunk things. Procedural data, after all, isn't just natural languagee: you got your part numbers, your geometries, your interwiring stuff, and lots more, all of them separately optimize-able.
The trap I always try and avoid: believing semantics are in some way an inherent or emergent property. Lots of times people in content systems have a near-religious belief in a sort of Platonic Semantics, which exists in this ether-space. Better to think of semantics as framework for NLP. My opinion, anyway. Keeps you out of a lot of trouble.
But what do I know? We could get yet another Mystery Emergent Behavior at ten trillion parameters in GPT-N that eliminates all or practically all problems with the solution, or somehow pre-guesses your future trips, or sees through your question(s) to the underlying problem. Right now, where I am, using any of this LLM stuff is strictly verboten, due to data restrictions and infosec. It's all after-hours projects at this point. So that's maybe another reason I'm thinking about smaller models - easier to get them past the gatekeepers.
I couldn't find the number with searches (I tried before) and I didn't even know that it might be helpful (didn't find it before, so how could I tell) but the AI spilled out and told me how useful it is.
This made me change my mind on search vs. AI. I got what I needed while I was unable to retrieve it with search. It was at least a 10x improvement over search.
But I do think there is space for both.
I don't really see ChatGPT as a search engine because I think that's not its strength, though it can certainly deliver in that area sometimes. I like using ChatGPT to clarify information I already have some idea about, and just need a straight forward summary for. This is different than asking it for specific shreds of information. If you treat ChatGPT as an intelligence and not a robot performing reflex over the internet, it's going to be correct often enough. But you always gotta use your head. I don't ask ChatGPT anything that's too far out of my domain of knowledge because, if it gets something wrong, I won't have the Spidey senses to realize it.
And rising.
What action would you recommend?
try yourself ask a random person the similarity between two sentences, without giving them a ruleset or examples.
Even with examples the result very from person to person.
I'm talking about a top physician who is of course a leader at using the AI to leverage his own advanced skills, along with contributing to the enhanced training of the model as they go along in a way where each physician's unique input can leverage each other's.
Man in the loop > Loop with no man
Black swan-type events are going to be hard to see even if you're looking for them. You can probably happen upon one by placing enough bets, but from a career/business perspective each bet can be pretty costly.
I've found that the semantic search things I've seen so far for local systems have been unusable, largely because of the time and cycles needed for the indexing.
But if that problem is ever fixed, I'll be all in on it.
Note that with transformer models the usual story is that people try a "zero-shot" approach to a task that works 70% of the time (e.g. the blog poster who asks ChatGPT one question and gets an incredible result isn't super-lucky) but when they go to a "few-shot" paradigm with a modest amount of training data (1000 samples, probably half a day of hardcore classifying... I've sustained much higher rates of image classification for a few days but found my visual system would start to malfunction)
By bet would be that it would make his work worse. Constraints are almost always what leads to great art. Unconstrained creative vision usually leads to the opposite.