cm2187

Well if you mandate DNA patching, how do you enforce it?

I agree, though one could make the argument that our modern nanny states have been pretty brutal at enforcing health policies during covid, and if they convince themselves that they can eradicate certain diseases by mandating DNA patching or pregnancy terminations, them doing so "for our own good" is in the realm of the possible.

But we are in coercion territory. What I am saying that we already practice eugenics without coercion, we just don't call it that.

The problem is "eugenics" has two meanings which is unhelpful for this discussion.

1) criminal practices of forced sterilisations, ethnic cleansing and mass assassinations to phase out undesired genes

2) the more generic practice of trying to improve the genetic characteristics of your children.

I don't think there is much point in debating 1). But we would be naive to think we are not already doing 2). What else is a prenatal test for down syndrome? What else is selecting your mating partner for desirable characteristics? In animals it's called breeding and it works pretty well. And if you can patch the DNA of your kids to remove potential risks of cancers or other deficiencies, why wouldn't you? Is it better to let cancer take its toll?

Same with job interviews. Right now Hr insists on us doing them over zoom so we get this absurd result that we eliminate candidates that perform too well to be true, at the risk of eliminating a genuine excellent candidate. You have to look a bit messy!

PC enthusiasts aren't exactly sentimental when in front of a spec sheet and a price list. Plus where else are you going to go. All manufacturers are hiking up their margin if you believe their stock prices.

Why? Part of the problem is that chip manufacturers (from tsmc to to memory makers) are reluctant to ramp up production as the AI bubble may pop and they would find themselves with huge over capacity, a scenario they have gone through many times.

By giving them stability of cash flows, the AI companies are enabling them to make those investments and to ramp up production. That's a good thing, not a bad thing. Over time it should ease the squeeze on chips.

The article says a large part is paid upfront.

In fact any application where the task is stable and the model good enough to address that task. As you suggest, industrial applications where a robot must deal with variants of the same repetitive task. Or a military drone which needs to be jamming proof.

Stupid question: I was under the impression that these models were trained on PB of data. Surely the amount of questions/response they can extract from querying a bigger model (Claude) is fairly modest. How is it not a drop vs the training dataset?

Random thought. Once models stabilise, could you possibly hardcode the model in gates? Or are they too large for a single chip?

But I understand the admin has gone up significantly. Though I presume AI is pretty good at generating the boiler plate bureaucratic work (privacy policy, anti slavery statements, etc).

Say what you want about their business practices, their products are top notch.

A bit premature I think. Right now it is employees making room for AI investments

I have news for you. That burger is the McDonald's commercial? It's most likely made out of plastic. That happy lottery winner? Probably a stock photo from one of the major visuals providers. And I am ready to bet my bankers don't have this hollywood white teeth looking of banks commercials. Since when is advertising real?

Today yes, but 40 years ago someone made the decision that a string was a char array and that every string manipulation going forward would require manipulating arrays. Talking about costly decisions.

It’s actually interesting to compare the pain and suffering of switching to a string datatype in the 80s (refactoring the limited code base then) vs the next 40 years of unnecessary boiler plate syntax and bugs for not having this type in key APIs.

A lot of pain and suffering to avoid having a string datatype.

The drive internally usually stores data in blocks larger than the 4k blocks it pretends to use to the OS (say it uses 16k blocks internally). So if your writes are sequential, it will most likely write the same 16k block in one go. If you write random 4k blocks, it will have to rewrite a full 16k block for every 4k block you write, hence write amplification.

So if you fill the drive over and over using sequential writes you can expect way more endurance than if you write random blocks.

Sequential vs random writes make a world of difference.

That’s why SSDs have overprovisioning, more so for enterprise drives. And why TRIM is important. That being said most workloads are sequential, particularly retail usage. Much easier to deal with for the controller than enterprise workload ratings that assume 4k random writes

Plus if AI is indeed a bubble, investors are already up to the neck in it. Think of all the key stocks of the S&P and Nasdaq that are lifted by AI: nvidia, microsoft, amazon, oracle, micron, intel, amd, etc. And add their lenders and the whole datacenter supply chain.

There are some european success stories in high speed rail too (France). China doesn’t let itself entangled in planning restrictions. US and European countries love over regulation and litigations. That’s a choice, and it comes with a steep cost.

VS the cost today yes. Nuclear wasn’t seen as a prohibitively expensive form of energy, quite the opposite. South korea and China are making it work. But if you build a new prototype every time and get bogged down in decades of environmental surveys, lawsuits and ever changing over regulations, any infrastructure project becomes prohibitive, whether it is nuclear, high speed rail, or anything else.

But if you define a new type, how is that breaking backward compatibility?

But what I don’t get reading the original article is that they present how to insert struct in an object oriented language as an intractable problem, whereas a good implementation with .net (as far as I can tell) has been out there for nearly 30 years. And C# was shameless about stealing from other languages.

I don’t think it can. Nuclear is cheap at scale, not when you build a single prototype every 20 years. You need large countries with large programmes to be successful.

I can't tell the difference between a 5 years old smartphone model and the latest generation.

There was a very interesting podcast that went into all the details of the AI supply chain shortage [1]

The key takeaway for smartphones isn't so much that iphones will cost $150-200 more, which apple customers have shown they can stomach. But that cheap $200 chinese smartphones will need have to hike prices by about the same amount, which will decimate that market.

[1] https://www.youtube.com/watch?v=mDG_Hx3BSUE

The computer is not exposed to the WAN (behind a firewall), the main way it could get infected is via a vulnerability in a browser, but these do get updated. And OS updates don't really protect you from malware in executables you install anyway.

I enjoy windows 10 hugely now that it is out of support. It became way better when microsoft started tormenting the users of win11 instead of win10, and now that windows update doesn't bring new catastrophes and unexpected reboot, the OS is finally not interfering with usage anymore.

How is that going to help him? "Our models are so inferior they are not deemed a threat unlike anthropic's"?

I think it is either a missile directed at anthropic, as retribution for not giving the DoD what it wants, in which case it is likely to resolve pretty quickly. Or it is a shift of policy toward export restrictions on powerful LLM and then every model will be impacted as they reach the threshold. In which case this could have massive implications of revenues, valuations, and the whole datacenter buildout. And frankly on the location of the white collar workforce if it is indeed a productivity multiplier, all countries reciprocate, and not all countries can match the US LLMs.

And why would the EU allow exports of chip manufacturing equipment if the US then restricts the export of derivatives of those chips to the EU?