It’s unwise to narrow the cause down to just one thing, and probably unfair to limit it to Twitter-related spiritual damage, but something is clearly not right with the richest and most powerful people in Silicon Valley. In some cases, this is pretty easy to identify; having a billion dollars and every material comfort imaginable […]
Everything I’ve seen about hyper-specific-specialty-data-trained, limited purpose AI - particularly for medical screenings - looks promising. The idea of a FOSS AI trained locally on my own collection of documents is intriguing, but I have a hard time believing that it will be useful beyond asking it to pull up any notes relevant to keywords of any given project I’m working on.
Conversely, nearly everything I’ve seen about general AI/LLMs, including my own (limited) experiences, provides little reason to believe the hype espoused by the folks promoting them.
welll, actually, llms are great at the exact things you just described… logic tree searches…
it just takes incredible computation and training at the moment.
if you filter your expectations to that limited scope, current ‘ai’ makes sense
but yeah, theres a ton of marketing pretending this is general ai, and its not even close.
I’ll be real with you - I was expecting a totally different experience than what happened when I saw the “welll, actually”, and I am happily surprised.
Agreed, they’re good tools for certain very specific tasks. Somehow my experience with them has produced the opposite of the Gell-Mann Amnesia effect, though. It’s hard for me to imagine that their output quality will exceed that of the 50^th percentile with the data collection and training methods in use.
Interacting with ChatGPT (or any LLM chatbot) without a goal is akin to buying a tool you have no projects for. And just as a hammer can’t provide plans for a grotto, chatbots can’t really talk me into there being a reason to chat.
I’d also be curious to train a local model on my past work, but it feels like it might keep my attention for a bit and then I’d run out of transformative ideas. At this point, I know this is the inescapable future, but I’m swinging wildly between thinking we’re going to find huge advantages and, well, apologist item descriptions.
My interactions with it were done to get a higher level of polish on some business related spreadsheets I was working on. I can manage certain tasks, but I don’t use those skills often enough to warrant scaling my knowledge and practice to that level. A buddy recommended using one to give me the necessary formulas, so I tried it out.
Not one suggestion from the bot worked, and as it apologized and offered other non solutions I noticed recurrences of previous answers and a degradation of the quality substantial enough for a relative novice like myself to spot the issues before implementation. Perhaps it will advance past this point, but I’m unsure that the people in charge of them are really the folks needed to actually hit that target.
Everything I’ve seen about hyper-specific-specialty-data-trained, limited purpose AI - particularly for medical screenings - looks promising. The idea of a FOSS AI trained locally on my own collection of documents is intriguing, but I have a hard time believing that it will be useful beyond asking it to pull up any notes relevant to keywords of any given project I’m working on.
Conversely, nearly everything I’ve seen about general AI/LLMs, including my own (limited) experiences, provides little reason to believe the hype espoused by the folks promoting them.
welll, actually, llms are great at the exact things you just described… logic tree searches… it just takes incredible computation and training at the moment.
if you filter your expectations to that limited scope, current ‘ai’ makes sense
but yeah, theres a ton of marketing pretending this is general ai, and its not even close.
I’ll be real with you - I was expecting a totally different experience than what happened when I saw the “welll, actually”, and I am happily surprised.
Agreed, they’re good tools for certain very specific tasks. Somehow my experience with them has produced the opposite of the Gell-Mann Amnesia effect, though. It’s hard for me to imagine that their output quality will exceed that of the 50^th percentile with the data collection and training methods in use.
Interacting with ChatGPT (or any LLM chatbot) without a goal is akin to buying a tool you have no projects for. And just as a hammer can’t provide plans for a grotto, chatbots can’t really talk me into there being a reason to chat.
I’d also be curious to train a local model on my past work, but it feels like it might keep my attention for a bit and then I’d run out of transformative ideas. At this point, I know this is the inescapable future, but I’m swinging wildly between thinking we’re going to find huge advantages and, well, apologist item descriptions.
My interactions with it were done to get a higher level of polish on some business related spreadsheets I was working on. I can manage certain tasks, but I don’t use those skills often enough to warrant scaling my knowledge and practice to that level. A buddy recommended using one to give me the necessary formulas, so I tried it out.
Not one suggestion from the bot worked, and as it apologized and offered other non solutions I noticed recurrences of previous answers and a degradation of the quality substantial enough for a relative novice like myself to spot the issues before implementation. Perhaps it will advance past this point, but I’m unsure that the people in charge of them are really the folks needed to actually hit that target.