HNHacker News
TopNewBestAskShowJobs

umutisik

417 karma · joined August 29, 2013

submissionscomments
umutisik··on Ask HN: Best books to understand semiconductor business?
+1 on Asianometry. It's more in depth than Chip War; in a much shorter amount of time.
umutisik··on Applied Category Theory Course
This agrees with my experience when I was a young mathematician looking for math that is useful.
umutisik··on The “Do Something About It” Club
It doesn’t need to be a club. You can do something about it as an individual, including recruiting others when needed.
umutisik··on Text classifiers are an underrated application of LLMs
And very likely this is coming to computer vision too with multi-modal GPT.
umutisik··on You're barely managing
I used to teach at a university, and am currently a manager at a tech company. It's interesting how similar these areas are in the sense that they are very hard to teach; and are currently mostly figured out from scratch from experience. The one teaching training that really worked for me was that we had a training where a more experienced peer recorded my lecture in a video, and then we watched it together. He did not give me feedback about what I was doing right or wrong, just asked me what I think about what I was seeing, and was essentially teaching me to self-observe. Perhaps there can be technologies to support this kind of self-observation on a day to day basis.
umutisik··on Refact Code LLM: 1.6B LLM for code that reaches 32% HumanEval
Phi-1-small also scores higher with 350M parameters. It helps to be specific about what the comparison is against when claiming SOTA.
umutisik··on Refact Code LLM: 1.6B LLM for code that reaches 32% HumanEval
The title is misleading This model is not "SOTA for the size", there are smaller models that do 10-18% better in absolute score. The text says it's SOTA "among similar models" where they probably compare with other models with permissive licensing.
umutisik··on Ask HN: Why did Python win?
I don't see this mentioned often: One thing that interpreted languages like Python have over compiled languages is that all libraries have to be open-source by default (unless they are wrappers). It makes it a lot easier to debug and put in minor fixes on the fly.
umutisik··on Artificial General Intelligence – A gentle introduction
We have the desire to continue existing. Hopefully we can build AGI systems in a way that makes them completely ok with being paused or deleted. Otherwise they will use up resources when not helpful to humans.
umutisik··on OpenWorm – A computational model of C. elegans worm
"Congratulations! You can now join the Devs project."
umutisik··on Ask HN: Am I hurting my programming career by not exploring AI coding tools?
Here's my experience so far:

ChatGPT saves time writing short functions maybe 10-20 lines. This is maybe 10 minutes a day.

ChatGPT helps start/debug things outside my domain where I would normally need to read some guides before even starting. This doesn't happen every week, but when it does, it's hours saved.

Copilot saves a bit of time with simple 1-3 line suggestions where you would already know how to write it, or can immediately verify correctness. This is probably 10 minutes per day saved as well.

So, on average, I get maybe 30 minutes or so per day saved with close to zero downside. It's not going to make or break a career but it's a good chunk of time.

umutisik··on Moving fast with the core Vim motions
If a Vim plugin could analyze what I do and suggest commands I don't already use that would save me time, that would be fantastic!
umutisik··on How much of AI's recent success is due to the Forer Effect?
The thing about the 'Forer Effect' is that it works both ways...
umutisik··on You and Your Research (1986)
What makes you excited about something in the case when it's not important?
umutisik··on Use Gröbner bases to solve polynomial equations
Gröbner basis methods do not scale well as the number of variables increases. The computational complexity becomes too high. Numerical Algebraic Geometry methods give the correct answer in many cases where there is a practical reason to solve the equations, but not a provably correct answer.
umutisik··on Category Theory ∩ Machine Learning
It is tempting to believe that category theory will shed new light on and simplify machine learning, just like it did in algebraic geometry, algebraic topology and other mathematical things. This is wishful thinking. Folks who care about doing something useful should stay away from this content.
umutisik··on Company, Team, Self
Put customers first. The company, team, and self (in a work context) are there to help customers.
umutisik··on Ask HN: Do you recall any book or course that made a topic finally click?
A Course in Mathematical Logic for Mathematicians, by Y. Manin https://math.uchicago.edu/~shmuel/lg-readings/Manin,%20Logic...
umutisik··on What will be the second order effects widespread AI?
Short term, a couple of things come to my mind:

Pretty soon, real ID will be needed to post / get visibility on most social media.

Productivity / efficiency gains in some white collar jobs like customer service. Fewer jobs but remaining ones will be better paid.

umutisik··on Ask HN: Tips to relearn how to care about my job?
I was in a similar situation 10 years ago, but in a less well-paid job :) I happened on a musician's talk about how to be a productive musician. He was saying that his most important suggestion was to "write everyday". It seemed obvious that musicians should write music everyday. But he meant writing your thoughts everyday on paper. (I am not a musician btw) I took that to heart and started writing in the mornings. The practice helped me to construct, then re-construct, my work and career from first principles.

That being said, "my current mental state is affecting my ability to learn effectively" sounds like you're in some distress. It will not hurt at all to seek help.

umutisik··on Poll: How are you doing currently?
Spending time in nature gets me out of the "I should be doing something now" mode you're talking about.
umutisik··on I draw figures for my mathematical lecture notes using Inkscape (2019)
I use a program called Latexdraw that gives the best of both worlds. You can draw with the GUI and get pstricks code and use latex expressions and math fonts. (Free, author accepts donations) http://latexdraw.sourceforge.net/
umutisik··on Ask HN: I got into MIT. Should I go?
It's not clear why Z needs to be undisclosed and it makes it harder to give you advice.

I second people talking about the 'regret minimization framework'. It sounds like you will be studying something technical. Look up big tech salaries on levels.fyi, there's no guarantee those numbers will be there but, 5-10 years into your career, you might be able to save your MIT tuition easily with a tech salary. If you got to that point, would you regret not going to MIT? On the other hand, if bootstrapping a start-up is important to you, it might be difficult to do that when you have college debt.

It's good that you're seeking advice. Writing down your thoughts clears the head too, and makes you notice spots where you're pushing to convince yourself. For big decisions like this, I write all my thoughts as I do my research. As I get closer to the decision, I make a list of pros, cons, known unknowns / risks, then sleep on it. I never regretted a decision I made in that way.

umutisik··on My First 80 Days of VR for Exercise
Exercise is the killer app for VR. If they figure out full body tracking too, they will give (and capture) a ton of value.
umutisik··on My First 80 Days of VR for Exercise
I had the same issue at first, but I was able to overcome it by being careful to not fully extend my elbows and using the jab less often.
umutisik··on Who is the most accurate world chess champion?
If “accuracy” measures how well a player matches computer chess, then as players continue to study more and more with chess programs, you would expect their play to match the programs more and more.

Personally I find it odd to measure how well the players match the computer program and call it accuracy. The computers do not open the game tree exhaustively so they give only one prediction of true min-max accuracy.

When Lee Sedol made move 78 in game 4 against AlphaGo, it reduced his accuracy but won him the game.

umutisik··on Let’s talk about founder compensation
I’m sure any replacement CEO would love a pay structure where the guaranteed comp is 25% but the bonus matches founder gains.
umutisik··on Turing Oversold?
Being the first one to claim a formal proof to a theorem is not the most important thing. What's really important is to prove things that improve our understanding, which both Goedel and Turing have done, in the fields of logic and computing respectively.
umutisik··on Safari 15 on Mac OS, a user interface mess
What I would like: Infinite tabs, they don’t get too small in the tab bar, newer ones to the right, older ones can be accessed by scrolling left. Older than last 15 don’t take up memory. Tabs get saved into disk/cloud and reappear when you restart the browser like on iOS. Basically so you can put off dealing with your open tabs indefinitely.
umutisik··on Crazy New Ideas
The essay seems to be about startups. It could use more examples to illustrate the point.
← PreviousPage 2 of 4Next →