On Sat, Jul 27, 2024 at 10:29:29AM -0700, PGC wrote:
> In Silicon Valley history, there are few spectacles as tantalizing—and as 
> perilous—as the sight of countless billions being funneled into a single 
> technological frontier. AI is, next to impressing us all, also starting to 
> show signs that it might just be the next bubble waiting to burst.
[...]
> 
> The parallels to past tech bubbles are striking, such as dot-com crash and 
> the autonomous driving hype of 2017. Capital continues to be vacuumed into 
> the AI sector with very little to no attention being paid to company 
> fundamentals.
> 
> What of the immediate financial hemorrhaging—such as OpenAI's projected $5 
> billion loss this year?
> 
> So, with all this in mind, is it still the right time to invest in a 
> cluster of Nvidia A100s and walk boldly into the future?

Depends. Do I have something that would require A100 in order to give me
results? Other than Chad Geppetto which is amusement, and not very
useful. I would not dare to have a database on my computer which
invents me non-existing results. If I had to install it, I would not
dare to use it. Chad Geppetto is, to me, software equivalent of
certain kind of people, who always are right, even if they have to lie
to prove it. Good for some entertainment. Not good for building
bridges, aeroplanes and medicines.

> To the computer and AI specialists reading this: Do you still see
> the incredible potential in large language models (LLMs) that
> justify the current hype and investment?

It looked like something big, until I learned about hallucination
without brakes. Until I have read about self driving car speeding up
to drive over a woman crossing the road. Ok, so you can teach it to
not drive over this one particular woman crossing this one particular
place. What about other situations, other people?

Maybe building bigger LLM is the answer or maybe not. There is no
proof, so far, but there is plenty of experimenting on live subjects.

> Has this technology made your lives easier, or made you 10x 
> more productive,

I have heard some people tried this. I believe it is akin to asking an
intern to writing me some initial code. After that I still need to
spend time looking for errors in the code and maybe also bending it
another way, because "intern" might not understood. Because Chad is
not equal to the person who sits in your business day and night, ten
years in a row.

It might work in some generic cases - something generic enough to not
require business specifics.

Oh and now they want Chad to remember. My questions would be
remembered. My questions may leak out. Something specific to what I
wanted to do may leak out... days or years later. No thanks. There is
going to be a whole insudrialt egaspione branch related to milking
Chad. People who used Chad's help when planning some future company
move may get sued for the leak or may have to prove they did not cause
the leak, resulting in smaller revenues.

> fulfilling Jensen Huang's prophecy that "People don't have 
> to learn programming anymore to develop software themselves"?

A long living phantasy, wet dream of people who will not learn to
program (and will not want to pay to those who want to learn,
maybe?). Look, we all have some specialty. Do I want to be a
neurosurgeon, with Chad Geppetto doing all dirty/bloody work in my
name and me getting the hard earned cash?

> And finally, where are all the killer apps built by hordes of agents
> conferring with each other and debugging and refining their code
> constructively?

Good question.

-- 
Regards,
Tomasz Rola

--
** A C programmer asked whether computer had Buddha's nature.      **
** As the answer, master did "rm -rif" on the programmer's home    **
** directory. And then the C programmer became enlightened...      **
**                                                                 **
** Tomasz Rola          mailto:[email protected]             **

-- 
You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/everything-list/Zrgoen%2BhFmYtA/ca%40tau1.ceti.pl.

Reply via email to