Up for banter around ML, Engineering and @Arsenal.
Palo Alto, CA
Joined September 2011
- Tweets507
- Following227
- Followers84
- Likes1.7K
gurkanwal singh brar retweeted
Markets excel at scaling a technology once its value becomes visible but are far less capable of sustaining an unfashionable idea through decades of uncertainty.
My argument for public support of curiosity-driven research:
outlookbusiness.com/magazine…
gurkanwal singh brar retweeted
If Tesla charges $1 per mile and it costs roughly 50 cents per mile, that would result in a gross margin of 50% and likely operating margin of around 35%. Hmm.. what does that sound like?
So, if Robotaxis are cheaper and a better experience than owning a car, what’s it worth?
Let’s do some quick maths.
gurkanwal singh brar retweeted
ਦੁਹਾਈ ਨੀ ਦੁਹਾਈ
ਨੈਣਾਂ ਨਾਲ ਹੋਣ ਲੱਗੀ ਨੈਣਾਂ ਦੀ ਲੜਾਈ
ਸੰਦਲੀ ਦੁਪੱਟੇ ਉਤੇ ਬੂਟੀਆਂ ਨੇ ਕਾਲੀਆਂ
ਕਿਥੇ ਕਿਥੇ ਖਹਿੰਦੀਆਂ ਨੇ ਕੰਨਾਂ ਦੀਆਂ ਵਾਲੀਆਂ
ਜ਼ੁਲਫ਼ਾਂ ਦੇ ਕੁੰਡਲੇ 'ਚ ਮੌਤ ਲਟਕਾਈ
ਦੁਹਾਈ ਨੀ ਦੁਹਾਈ
-- ਨੰਦ ਲਾਲ ਨੂਰਪੁਰੀ
gurkanwal singh brar retweeted
Yogesh and I are excited to announce we have raised a $2.8M pre-seed round! The last 8 months have been an amazing adventure, but there’s so much more to come. We are expanding the team, building AI training infra that’s never been built before, and launching our protocol 👇
We’re happy to announce that we raised a $2.8M Pre-Seed round! Investors include @LongHashVC, @_inceptioncap, @protocollabs, @trgcapi, @l2iterative, Zero Knowledge, @fenbushi, Quaker Capital, @OrangeDAOxyz, @CapitalZephyrus, and @EV3ventures! 👇
gurkanwal singh brar retweeted
Aiming to automate the tedious aspects of digital communication in the workplace @querypal!
Check out our new blog:
querypal.com/digital-workspa…
AI can automate software engineering but not engineering. Learn to build and treat everything as a tool - software today and ml agents tomorrow. You’ve got a job for life.
Excellent post @dwarkesh_sp better than most review papers out there.
I think a missing piece in self play is the construction of conventional testing tools and digital twins or simulated environments within which the virtuous loop of (LLM Agent) -Action-> Simulation -NextState-> Verifier -Reward->(LLM Agent) can run.
To that end, LLM's can help accelerate the development of Simulators, this process is easiest to see in the FunSearch paper which combines Evolution with conventional unit testing.
Adding Human in the loop above will also help in approaching novel discoveries, This will be slower but will compound imo.
A graph theoretic view on my inability to write creatively in my 30's. Graph of concerns grows exponentially with age, while poetry requires zooming into a subgraph, so it fills your consciousness enabling extrapolation. More on this here:
open.substack.com/pub/gkbrar…
gurkanwal singh brar retweeted
The general architecture used in this paper: HumanWrittenSpec -> Evolve using LLM -> HumanWrittenTest-ScoreFunc is amazing leverage across engineering teams. Best Engineers will be the ones who can up-level into Senior mindsets to focus on Spec, Test before implementation. So long leetcode grinding.
gurkanwal singh brar retweeted
Announcing the 10th #LoFi (#LocalFirst) meetup on Tuesday, November 28, at 11:00 AM EST. We have another fantastic lineup featuring @gsvarovsky, @threepointone, and @RangerMauve. Make sure to tune in by visiting lofi.software.
I believe LLMs are in their x86 fixes everything phase with LLMs and RAG. This will serve most general use cases well into the near future like CPUs have.
But i foresee a future where many small
heterogeneous LLMs work together like modern day SOCs. More on this here:
gkbrar.substack.com/p/coming…
and given LLMs are rushing through their evolution at crazy pace, the heterogeneous future can’t be that far off.
Where heterogeneity was a result of power, performance trade offs for chips, in LLMs it’s a result of Accuracy, Generality and Cost to an extent.
gurkanwal singh brar retweeted
“My belief is that all work in the enterprise, and especially for DevOps engineers, is about communication,” Dev Nag @devnag told @thenewstack. “Not just communication between people, but communication between people and machines, too.”
thenewstack.io/how-generativ…
#genai #devops #SRE
gurkanwal singh brar retweeted
It's been a while since I posted about what I'm up to.
At Luda, we’re building a new medium for play based on R&D started in Google Research five years ago. It’s called Mels.
Excellent writeup.
Here is how i parsed it: Planning in LLMs works for abstract problems which require no sub goal implementation - which is why single shot accurate code generation is still an unsolved problem. And also why interfaces with human in the loop will win.
#LLM #GenerativeAI #ChatGPT
📢 My new piece for @blogCACM summarizing my views and our ongoing work on the planning/reasoning abilities of LLMs..
cacm.acm.org/blogs/blog-cacm…