Raghuveer Varahagiri

8 Followers
22 Following
15 Posts

Brain + Brain-Computer-Interface + GenerativeAI = endless possibilities.

The brain need not be a human brain.

How long before we humans are competing with our animal (or plant?) friends for basic logical and creative jobs.

#GenerativeAI #BCI #LLMs

Let's give the nuclear codes to GPT.

#MoveFast

Next idea: Ensemble LLMs - use multiple LLMs to generate against the same prompt and synthesize a more robust final response.

Maybe each model can engage in a Socratic debate with the others in the ensemble to critically question and bolster or weaken viewpoints.

#LLMs #GenerativeAI

Prediction:

Once we have Diffusion Models that can run locally on a smartphone, we'll see an explosion of consumer use cases. Some can take off even with models running elsewhere, over the network, but in near real-time.

E.g, Live interactive SUPER-RESOLUTION on mobile cameras

@esh

A simple thumbrule that could help everyone if AWS agrees and cooperates --

everything launched or promoted at re:Invent is a AWS-${thing}

Hands-on AWS | All things AWS, Cloud, Data and Digital

AWS Certification Exams: Upcoming Changes

PSI is no longer an option after tomorrow for AWS exams. Pearson VUE is the sole exam provider.

AWS Certified Developer Associate and DevOps Professional Exams are getting an update in early 2023.

#aws #certification

https://shorts.handsonaws.com/post/aws-and-databricks-among-many-leaders-in-the-gartner-magic-quadrant-for-2022-cloud-dbms/

Hands-on AWS | All things AWS, Cloud, Data and Digital

The volume of AI-generated data and content is going to explode.

Future models will face an uphill challenge in curating training data and filtering it out.

Indiscriminately trained models of tomorrow might pickup on subtle patterns in AI output of today.

#chatgpt #data #ai

Emerging concern:

DATA POLLUTION.

AI-generated content in various domains is no-doubt impressive and useful. It is also often flawed - sometimes in obvious ways, sometimes in very subtle ways.

We can improve on flaws we can see. Risk lies with the imperceptible and insidious.

#chatgpt #data #ai

Can there be a Large Brain Model that models not just (written/spoken) language but all patterns of neural activity at a massive scale -- thanks to widely deployed Brain-Computer interfaces or implants.

So the model could "auto complete" a thought / feeling / action?

Dystopia?

#chatgpt #ai