Huh? Market cap is the standard, recognized way of measuring the size of companies.
Sure you could measure land owned, or employee count, or sales, or a million other things.
But market cap is the actual value of the company, which serves as a way to account for and weight all of the factors that matter in terms of a company's health and scale.
Generative AI is going to be as big of a shift as the internet. We already have chatbots that have internalized the entire internet of billions of humans. We now have the ability to have an Einstein-level of genius in our pockets, and we can ask endless questions. Soon, we’ll have personal AI doctors, lawyers, chefs, tutors, teachers, artists, you name it. Every single industry will have personal assistants that will not hallucinate because they are strictly being trained on data in their particular fields.
Training on data in a particular field will not prevent hallucinations, because the generation algorithm itself works on hallucinations. Anything below the resolution of “looks like the source” is indistinguishable.
When you ask it to do arithmetic operations, its approximation performance is surprisingly good. The expectation is that its responses are good approximations of the query, but not necessarily perfectly accurate. The problem though is that good approximations in the English text domain might care more about good presentation than accurate information content.
the container ship that was used to deliver the dildo in your nightstand drawer all the way from China to the US had probably emitted more carbon dioxide than training GPT4 did.
I think it makes no sense to foam at the mouth about AI energy consumption when it's a drop in an Olympic-size swimming pool.
I also think it's just an angle for you people to attack a technology you dislike. just like anti-car zealots don't give a shit whether a car is ICE or EV and have seamlessly switched to preaching about 'walkable cities' instead of their previous rhetoric, you'll adopt some other rhetoric even when the models no longer require whole datacenters to train.
> Every single industry will have personal assistants that will not hallucinate because they are strictly being trained on data in their particular fields.
Every single output from an LLM is a hallucination. It's just that for good models, most of them make sense.
I think that generative AI is a much bigger deal than the internet and maybe even the Industrial Revolution. The internet mainly sped up existing processes and made them accessible to everyone. We’ve had mail order for over a century, transatlantic communication through the telegraph and radio, and the printing press.
If the Industrial Revolution was so impactful because it boosted human productivity, then the ability to spawn hundreds, even thousands, of intelligent 'humans' at will must be an even bigger deal.
I'm not that impressed until one of those "intelligent humans" can make me a sandwich and deliver it to me. Gen AI's productivity boosting capabilities are seriously limited outside office work, and even there it can't be trusted to perform totally autonomously. It's a big deal, but not a revolution.
Now, if combine robotics with Gen AI and some other solutions to form true AGI that can do both tasks in physical world and virtual, then things will get more interesting.
> We now have the ability to have an Einstein-level of genius in our pockets
It's pretty cool but what you really mean is that it's a better UI to an encyclopedia and a story / art generator. It has no ability to reason at all and often struggles with simple things.
As of now all your chatbots give me the exact answer, which made me reach out to the chatbot, and most of the times I was forced to reach out to the chatbot. Plus companies have weird excuse not to hire humans dealing with my ticket.
28 comments
[ 3.8 ms ] story [ 81.4 ms ] threadSure you could measure land owned, or employee count, or sales, or a million other things.
But market cap is the actual value of the company, which serves as a way to account for and weight all of the factors that matter in terms of a company's health and scale.
And a genius like chatbot is useless if it's used for mindless and malicious tasks.
The genius is also still in question. Einstein expressed something new and didn't just repeat existing knowledge.
I also think it's just an angle for you people to attack a technology you dislike. just like anti-car zealots don't give a shit whether a car is ICE or EV and have seamlessly switched to preaching about 'walkable cities' instead of their previous rhetoric, you'll adopt some other rhetoric even when the models no longer require whole datacenters to train.
Every single output from an LLM is a hallucination. It's just that for good models, most of them make sense.
Yes it's always hallucinating. The hallucination just often happens to be the answer you were looking for.
If the Industrial Revolution was so impactful because it boosted human productivity, then the ability to spawn hundreds, even thousands, of intelligent 'humans' at will must be an even bigger deal.
Now, if combine robotics with Gen AI and some other solutions to form true AGI that can do both tasks in physical world and virtual, then things will get more interesting.
It's pretty cool but what you really mean is that it's a better UI to an encyclopedia and a story / art generator. It has no ability to reason at all and often struggles with simple things.
It remains to be seen if this can be solved in a generalizable way.