HNHacker News
TopNewBestAskShowJobs

abm53

165 karma · joined January 12, 2020

submissionscomments
abm53··on Chemistry Nobel: Computational protein design and protein structure prediction
I did raise an eyebrow at it too, but I doubt his contribution was entirely “management of resources”.

I think one must also give him the credit for the vision, risk taking and drive to apply the resources at his disposal, and RL, to these particular problems.

Without that push this research would never have been done, but there may have been many fungible people willing to iron out the details (and, to be fair, contribute some important ideas along the way).

I’m not a proponent of the “great man” theory of history, but based on the above I can see that this could be fair (although I have no way of telling if internally this is actually how it played out).

abm53··on TSMC execs allegedly dismissed OpenAI CEO Sam Altman as 'podcasting bro'
I think this illustrates the biggest failure mode when people start using LLMs: asking it to do too much in one step.

It’s a very useful tool, not magic.

abm53··on TSMC execs allegedly dismissed OpenAI CEO Sam Altman as 'podcasting bro'
And I would go further… these “common tasks” cover 80% of the work in even the most demanding engineering or research positions.
abm53··on US judge says FTX can sell its Anthropic shares; FTX invested $500M in 2021
> The hedge funds (usually) discount the expected value by a ton when deciding how much to pay.

That’s called a risk premium.

abm53··on Is it insider trading if I bought Boeing puts while inside the wrecked airplane?
It’s often a dangerous kind of answer.

I once saw somebody say something to the effect that in every Hacker News thread, there is always a highly upvoted comment that sounds completely plausible, well argued, made by somebody who appears highly qualified to answer, and that is completely incorrect.

I don’t think it’s always true, but it often is.

abm53··on Training for one trillion parameter model backed by Intel and US govt has begun
In the space of possibilities, the volume occupied by “not quite smart enough to outwit a corporation but smart enough to outwit a government” seems so vanishingly small as to almost be disregarded entirely.
abm53··on We have reached an agreement in principle for Sam to return to OpenAI as CEO
I think another factor is that they had very limited time. It was clear they needed to pick a side and build momentum quickly.

They couldn’t sit back and dwell on it for a few days because then the decision (i.e. the status quo) would have been made for them.

abm53··on Binance founder Changpeng Zhao agrees to step down, plead guilty
I use a Google mesh network and get the same issue, which I assume is related to it using 8.8.8.8 by default.
abm53··on Sam Altman, Greg Brockman and others to join Microsoft
There is a third option where he stayed, they managed to find a compromise, and in so doing kept their influence in the space to a large extent.
abm53··on I disagree with Geoff Hinton regarding "glorified autocomplete"
You can ask ChatGPT to solve maths problems which are not in its training data, and it will answer an astonishing amount of them correctly.

The fact that we have trained it on examples of human-produced maths texts (rather than through interacting with the world over several millennia) seems like more of an implementation detail and not piece of evidence about whether it has “understood” or not.

abm53··on Three senior researchers have resigned from OpenAI
If you are building something that is end-user facing that relies on ChatGPT then that was always a huge and risky bet on the future of OpenAI.

In addition, it would likely be some time, possibly years, before it would be ready for production.

Perhaps recent events have just brought that more clearly into focus for you.

abm53··on Google Brain founder says big tech is lying about AI danger
An alternative idea to the regulatory moat thesis is that it serves Big Tech’s interests to have people think it is dangerous because then surely it must also be incredibly valuable (and hence lead to high Big Tech valuations).

I think it was Cory Doctorow who first pointed this out.

abm53··on Work only 3 hours a day, but everyday (2016)
For me the open question is if you decided up-front to only work 3–4 hours, would you be as productive as you are in those 3–4 hours out of an 8 hour day?

Highly job and person dependent I suspect.

abm53··on AI hype is built on flawed test scores
Perhaps they are trying to help people get the best out of a tool which they themselves find very useful?
abm53··on LLMs can't self-correct in reasoning tasks, DeepMind study finds
These models produce correct answers to many problems that require “reasoning” (for any sensible meaning of the word), that are not in their training set.
abm53··on Blackmagic Camera for iPhone
I’ve seen enough videos with otherwise high production values to make me suspect there is a valid trade-off to keeping autofocus on.
abm53··on Does Market Timing Work?
Sorry, then yes to answer your actual question, I don’t think that would commonly be called “market timing” (although of course the phrase could be an apt description of that strategy).

All the uses of that expression that I see refer to an investor whose principal aim is to buy-and-hold to capture beta, but simply wants to try and pick the right moment.

abm53··on Does Market Timing Work?
How do you know that you’re currently at a local low/high?
abm53··on The English cottage where John le Carré wrote Tinker Tailor Soldier Spy
I enjoy this version, but can’t help feeling that Gary Oldman is miscast and the wardrobe wrong.
abm53··on The concurrence of three climatic events
The article is mostly about current conditions, not predictions.
abm53··on America will soon see a wave of bank mergers?
The problem is that bank lending is the main form of money creation in the economy.

If retail banks switched en mass to backing deposits 1-to-1 with Fed reserves then the Fed would quickly take strong action to disincentivise this strongly contractionary effect.

They have plenty of legal tools to do this (e.g. IORB rate). They don’t need to resort to not giving out licenses.

abm53··on Geoffrey Hinton leaves Google and warns of danger ahead
He partly answers this in the article: “because if I didn’t, someone else would”.

He states himself that it’s not a convincing argument to some.

But it surely carries some weight: in developing nuclear weapons many scientists made the same calculation even though the invention is a wicked one, in and of itself.

abm53··on Credit Suisse sheds nearly 25%, key backer says no more money
Can you explain what you mean by “assume all their liabilities” in practical terms?
abm53··on FDIC Takes over Silicon Valley Bank
If that was the market clearing price, then to the extent that rate expectations set the bond price (which of course is debatable) this would have been the consensus view at the time.
abm53··on Andreessen Horowitz and 'zero interest rate phenomena'
And that cost is taken into account when people value the merits of a particular investment/company.

If bonds yield 100% p/a then an investment with even a decent potential payout in 10 years time will be seen as worthless.

But one that is likely to pay out a dividend tomorrow may not be.

That’s the whole premise behind discounted cashflow analysis: mechanistically taking into account the opportunity cost of the risk-free interest rate to derive a valuation.

abm53··on Andreessen Horowitz and 'zero interest rate phenomena'
The difference is that the bond return is (to first order) riskless and risk has a cost.

What is the variance on the VC return (and is $9 a realistic mean of the distribution to begin with?).

abm53··on Microsoft's AI Bing also generated factual errors at launch
Right, it’s (very roughly speaking) an approximation to an algorithm which maximises the probability of the entire response by factorising the distribution as a sequence of conditional distributions.

If the approximation works, and the training data is coherent, it will produce coherent responses.

The surprising thing to me is that this does work so much of the time.

abm53··on Microsoft's AI Bing also generated factual errors at launch
I think one of the clear lessons of ChatGPT is that the line between word prediction and intelligence (for some definition) is somewhat blurry.
abm53··on Google employees criticize CEO for “dumpster fire” response to ChatGPT
Often what you’d like to do has not been asked for exactly before.

And almost all programmers need to do something quick in a language or framework they aren’t experts in.

ChatGPT can be extremely useful in those cases in my experience.

abm53··on Stripe increases price for business in the European Economic Area
On the other hand, transaction networks clearly add massive value, so it’s very likely that net of fees (without which the network may not exist at all) these networks still add value.

So it’s more a question of finding the optimal equilibrium rather than viewing fees as “deadweight”.

← PreviousPage 2 of 3Next →