Sorry if this is a duplicate. Deployment schedules, data centres, that type of stuff - do we have a view?

Hi all.

My journey in this space has been interesting, i got hooked on GPT-3 back in that beta stage, the writer hadn’t multiplied like pigeon dung yet, (sorry)… and it was pretty trivial to get something going but these guys this weird paranoia about the “dangers” of this system all a little unclear as to the how. Now i’m from that world, build - deliver - NEXT , i dont do cradle to grave one service type stuff. I derive satisfaction from the over the line … in their hands, they happy stuff.

Apologies, this is turning into a wffle fest. I posit that Enterprise and comercial devolpers do struggle significantly in this space in ways that they may find difficult to …get a hold of and verbalise. projection is obviously swithed to high right now.

and i think i’ve worked it out. Our world is a little different to this space - and i think we miss it because we don’t recognize the academia aspect of it.

Like Arvix… thats the stuff that normally comes wuith a press release, fancy grqaphics, sme training, launch party etc. so how do you keep current - thats stuff over there. okay, great - so whats this thing? … read the short bit and then if its relevant read the paper.

yes, see if you tell me there is a high risk with piece of software , i may sniff at it but then i need to get permission from you for what? no this, this, this , this… okay, “thinks WTF”, then wait a moment 5 uers testing and its same as release… REALLY!, 5 users testing.?

so just a culture quiet alien … and yeah i think i get it, Academics - data scientists, ML engineers and maths geeks (i mean that in the nicest way)

so i experience the texure of model fluctuating at different times , it like its a new person/attitude, soryy i am anthropomorphizer, guilty. get over its aractical aproach to help joe citizen break into the prompting space etc in my opinion.

so commerce is different. screaming big risk because a model may say something somebody finds offensive - probability quite , outside world is quite as dainty or infantile as that. SO thats what scared you guys… fake news, hallucinations and whatever.

okay, nice. risk + project = not a god mix. the fact that people experience that type of stuff right in their mailbox evry day on a scale that dwarfs the arb sydney etc , hysterical bs - i mean seriously risk to who and what?

so i wasnt convinced that the microsft thing was going to be wine and roses, the invest early then use the third party to musle up with all key aquistions in the domains is not new. However there is the openai corporate structure that may be a problem , and the we see Orca announcements and the dual track what is who thing becomes odd - the microsft forms to request acces - i just couldn’t , been deling with for decades, have acccess directly and now you an entirely whole new set of docs far behind what we have in place already, odd.

And roadshow - preaching that your product may kill the world , HOW? Look i love the space . and i will suffer the entire religion change python, apache , linu x etc.

But i’m starting to wonder are we certain the product is across the chasm b ecause … it doesn’t walk on water, iwould expect a would would need targetting to begin an autonomous agent mas hack say. and just like every other tool , weapon … vehicle. could there be scenarios where the product killed people ? yes… so what should we do… we i think we take a proactive approach mitigate aas much as we can , then insure against the other larger perhaps risks. great product out the door …

i see models changing and i try to finds patterns and i start thinking that the time of day has an effect, is that load ., is that a different cluster…what… you dont to build the monkey for your back. you want to be able to provide your value add predictably to your market s they like it and when.

is it too early to be playing in this pool perhaps. everybody is busy freaking out the potential market with yeah scifi movie scenarios. you know, da,mn those were such aproblem to stop from killing pedestyriands especiall when we parked them on a hill. you think we’l get sued for that… hmmm, you bet, we better terrify any potentiual customer with a barrage of risk notices before we hide it in the contract they sign.

I feel as if the is text generators each beautiful in there own unique way. but generators will generate . The stuff that we really need to fear - will come out of military labs, and please the horse has bolted long already, you couldn’t stop this with all the archangels in tow, its out there , done. and all things duly considered - yes generative AI has been a mind blow. but you do know in 18 months it’ll be something else.

so the moving past the chasing the AGI model obsession and the deliver to market, put infrastructure in place - ealy adopters , ealry majorty schlep… i dont think we are there. the models are out, they are product or isn’t that true. is this actually the sidehustle for the apollo team to land AGI.

sorry, i read some of the posts in the forum and i think some AI fatigue may be setting sorry about that turning into a rant. i hope there was something useful to someone in there.

OpenAi , kind of feeling like this stuff may be the side hustle to keep the apollo team going to create the ultimate …AGI. APIs were feantastic way of surfacing the product - you got a bunch playing just because they like to, It does not walk on water… and when a baby goes to the toilet on your lounge carpet, it can even be cute - a teenager?, now that a whole new ball game - you guys are living in AI time … and its incredible how quickly things are commoditised. i would be testing your market as it almost feels like honeymoon is over and its hard to miss that fame is having an affect.

i know i will regret creating the the topic. I’m incredibly excited about the space , so its not meant as a tear down, well… i am sick of the hysteria in particular at the moment. Get an agent flow that can focus long to take over the world , or cure the models alzhiemers byut please bad will do stuff with anything, so draw a line from GPT-4 to extenction else leave it alone or at least not the megaphone, get solid with party line and stop carving capabilities out of the models to mitigate some fringe that you’l nebversee above the scammers hitting you in the face. do not infantilise the entire specie , you may be surprised at what people have said to them and really the MDs and the shrinks they are all perfect, no funny advice por anything there - how could it, its a human (and at least wont get sued)… peoplle sometimes recognizeBS , more than you may think.

if you suffered it to here - sorry :wink:

1 Like

Complexity is evil. The more complex and the more things you combine the scarier it looks. The fear of the new.

It’s about the same as pressing a button to send hundrets of rockets to destroy the world. Let’s all just believe that this is all too complicated for evil people.

I bet there are many out there who find something and just stop working on it the minute they see it could potentially destroy the world.