Abinav R (@AbinavRavi) profile picture

ML Engineer @ACVAuctions. Personal Musings on https://teapawt.net/t.co/UNliqzES0o

Chennai, India
Joined July 2017
Its that time of the year and I have been examining what I have been doing with my time. Apparently I waste a lot of it and give very convincing excuses to self. I am planning to do some changes over the next 4 weeks to see if it's worth considering them for a resolution.
1
13
Time to understand databases slowly.
2
3
99
While most might think it is pointless to code by hand I am going to continue to do this for my personal projects for the sake of learning. I really want that struggle. Productivity is not everything when it's personal. #SoftwareDevelopment #programming
1
4
192
The success of JEV basically tells me that we are going to build task specific general purpose models. LLMs will be used for generation tasks and JEV like models will be used for discriminative tasks.
1
27
There will come a time when a single provider will build both that work together so well that it might become a package deal to sell enterprises in the name of reliability and safety.
5
Been spending sometime diving into ML inference engineering recently.Wrote a breakdown on dynamic batching—specifically looking at how to aggregate payloads for fixed-tensor inputs (like images) to max out GPU throughput Check out the full post here: abinavravi.github.io/blogs/d…
1
1
142
Abinav R retweeted
Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.
3,372
12,211
2,709
71,664
9,650,914
Abinav R retweeted
We all can agree @stanfordnlp is far ahead of most other academic labs when it comes to modern frontier AI. However, a fair question to ask is how can we entrust Stanford NLP as this sole arbiter role when most of its students and faculty have deep financial entanglements with the frontier companies they will need to evaluate either directly and indirectly (via shared investments)? This makes them anything but independent. Further, the choice of how we build a future truly independent third-party evaluator is so critical that we should have any single university or non-profit, including non-Stanford groups, in full control as it creates a single point of failure. Any such body should be a constellation of centers of excellence (COE) spread across the country to capture diverse perspectives and interests — not just the Valley — holding each other accountable. StanfordNLP will, without question, be a leading COE. Further, a tangential, but important, reason to have a multi-COE framework is them to uplift each other so we make our academic groups robust nationwide in this new post-intelligence era. Stanford NLP will, no doubt, also have a role to play there.
I propose Stanford NLP as an independent third-party evaluator under @DarioAmodei’s 3 step plan. For important parts of the work, universities would be better than any other organization (see below 🧵👇), and, of university groups, @stanfordnlp would be the best one to choose. 😊
11
16
2
163
19,579
The craft of software engineering is going to become more niche. People who invest in learning the craft will be treated like artists. Everyone can paint doesn't mean they become an artist.
2
88
The tribal need for stories and good companionship never really goes away. It shows up in moments of anxiety as relief and that is when you realise though there is evolution in personalized content nothing really beats sharing a meal and stories with friends.
2
20
Agentic coding is feeding the Dunning-Kruger loop. When AI writes 100% of your code, you get the illusion of competence without the actual skill. Until recently I was only aware of an old pattern used on startup events in fastapi and recently the docs helped me learn the new one.
2
113
I was having a very powerful conversation with a leader in my company and he made me realise while AI tools are great in accomplishing tasks the power lies in soft skills and navigating stuff around tasks.
2
38
With so much FOMO around the one thing that I have realised that I can do is keep learning new things consistently . We are in a new age and by consistently shifting priorities and focus we ll just burn out.
2
39
abinavravi.github.io/softwar… wrote a blog on a book that I read and how it relates to the work we do and why AI coding hasn't given returns in enterprises as expected.
1
3
57
Been revisiting some concepts in ML. Class imbalance is one such concept which we come across everyday. Basically class imbalance is when one category of data has more samples in order of magnitude and this skews what could be learnt.
1
3
305
Was reading about feature engineering which used to be an important art but with deep learning methods are dying a slow death. Feature engineering often involves dealing with missing data.
1
4
Missing data can be classified into 3 types 1. Missing not at Random 2. Missing at Random 3. Missing completely at Random For missing at Random we shouldn't delete rows and for missing not at Random we shouldn't delete columns. Completely random can be deleted.
1