Yeshh. 🏺

research at , building , trying to write stuff!

[Yeshh. 🏺 @Yeshhaha]

I just had an interesting observation.
- when using claude, i don't care to elaborate things i trust that it'll use common sense and rely on context
- when using chatGPT, i make sure that i say use the context, understand the boundaries and elaborate
- when I'm using gemini, i talk to it like a little kid, even within the same chat, i find myself abbreviating things or giving very detailed instructions.

i think i'm just implementing subconscious lessons here that can act as proxies for daily use efficiency, at least for me

#chatGPT #claude #gemini #AI

[Yeshh. 🏺 @Yeshhaha]

@ryanbrewer To put it simply, I think most are just busy making llms edit pictures or rewrite emails to look beyond what they can actually do.

[Yeshh. 🏺 @Yeshhaha]

It’s like this: as humans consume content from other humans which is produced/augumented/assisted by AI, whose thoughts or world view are we subscribing to.

- is it that persons cause they cared to write about it
- is it the AIs
- is it just humanity’s, since that’s where AI gets its words from.

[Yeshh. 🏺 @Yeshhaha]

I think I’ve just consumed too much information on pro ai and against ai for content production and writing.

I don’t think I have an opinion for pro and against, i wouldn’t say I’m confused, I just see both the sides now.

I gotta write about this.
#AI #writing

[Yeshh. 🏺 @Yeshhaha]

@ashitapatodia It very much depends on what your goals are. Whoop is great at tracking some things and not so great at some others, but without goals or improvement plans, it’ll just be a lot of data that might end up making you feel guilty

[Yeshh. 🏺 @Yeshhaha]

@ashleymayer “You become what you consume”, someone said.

[Yeshh. 🏺 @Yeshhaha]

@RichDevLab @trikcode I think that’s what karpathy wrote about a while ago. In the future it’s less about AI availability and literacy but more about affordability.

"...Their child will be tutored by GPT-8-pro-max-high, yours by GPT-6 mini.”

https://t.co/Xk1MsgTW6i

[Andrej Karpathy @karpathy]

https://t.co/3CR1HXfFxb

[Yeshh. 🏺 @Yeshhaha]

@AishwaryaDevv @trikcode Open models are alright.
But closed models have a lot more at stake and thus the vigor with which companies, labs and governments will make them grow cannot be compared with open models.

[Yeshh. 🏺 @Yeshhaha]

@trikcode This also points out the necessity for each country to have and encourage their own foundational AI research.

I think going on governmental control will play a much more central role in AI deployment and availability. As we saw with what USG wanted to do with Fable.

[Yeshh. 🏺 @Yeshhaha]

My entire twitter thread right now is people fawning over fable. So much excitement, as though we’re all getting back a lost twin. So the point still stands about model excitement!

[Yeshh. 🏺 @Yeshhaha]

Prof. Umbridges line 'progress for progresses sake must be discouraged', feels so relevant after reading this.

My understanding of the following article: deciding AI Expenditure is very reliant on chip design, cost and life cycle. But the very progress in those technologies is increasing expenditure rapidly.

Let's assume A builds chips, B builds data centers and rents out compute, C purchases compute. B should have invested a significant capital for the first round of chips, and C gets them because they provide the best cost for usage and inherently the best revenue/profits.

But when A comes up with a exponentially better chip within a year of the previous one, and D as a data center offers it, C would obviously choose D as this one just financially better. Now B has to either hope someone chooses them, or eat up the cost of the previous chips and update to the new ones. Increasing total AI market expenditure.

I think this analogy summarises what i learnt from that article and it seems so unfair and inefficient. I'm not sure if I fully understand it and I'm pretty sure there's definitely nuances to interaction, but with all the waves NVDIA creates at their expos, I can't help believe that this is very true.

https://t.co/YGDIhUVna3

#AI #PublicPolicy #ChipDesign #RAI #GoldmanSachs

[Yeshh. 🏺 @Yeshhaha]

@josephradhik One of my goals in life is for people to miss my cooking.
They should be like, ‘oh, man Yesh cooks this sooooo good, I wish he was here to make it again’.

[Yeshh. 🏺 @Yeshhaha]

@BenjaminDEKR Well, you atleast got a response. Mine just disappeared.

[Yeshh. 🏺 @Yeshhaha]

Reading about creative resistance to AI. Really fascinating.

Would be even fascinating to see if anyone has looked at the differences in resistance to AI, from a religious lens.

I’m gonna ask Fable to do it for me. 🤯

#AI #Creatives #AIPolicy

[Yeshh. 🏺 @Yeshhaha]

@marconoiz why?

[Yeshh. 🏺 @Yeshhaha]

@0xsachi i agree with a lot of others in this thread, if agi is supposed to be owned, it doesn't make sense nor it is fair.
regulation, laws and rules make sense, but ownership shouldn't be the way for agi.

[Yeshh. 🏺 @Yeshhaha]

going back to the original thread, i'm still deciding on a methodology, although mixed methods seems obvious here, computational analysis also seems to have its merits. particularly since @X made their api pay per use.

standby for more updates.

[Yeshh. 🏺 @Yeshhaha]

in the last three days of my break from here, i was trying to design a new study to understand collective anxiety in AI, from an Indian perspective.

There are quite a few tangential studies looking about the WEIRD population and their anxiety/collective anxiety about AI. But I couldn't find any relevant ones for south asia, specifically India.

#CollectiveAnxiety #AI #Research #PatronusLabs

[Yeshh. 🏺 @Yeshhaha]

By the gap I mean the gap in behavioural research, particularly in areas such as human-ai interaction, ai trust research or artificial intimacy.

I'm also looking to talk to people working on problems in these areas, scale doesn't matter, i wish to have conversations.

#AI #research #behavioural_research #india #ProblemSolving