janekm

12 hours on a single CPU core, or on a GPU?

The advice from my biology teacher in the 80s is still just as valid today (eat more fruits and vegetables and whole grains, eat less processed foods and sugar and saturated fats). See the recommendations from the AHA for example: https://www.ahajournals.org/doi/10.1161/CIR.0000000000001031

That consensus dietary recommendations change frequently is itself a myth.

Your proposed diet matches very closely to official dietary guidelines, such as the ones from the AHA: https://www.ahajournals.org/doi/10.1161/CIR.0000000000001031

(1) adjust energy intake and expenditure to achieve and maintain a healthy body weight; (2) eat plenty and a variety of fruits and vegetables; (3) choose whole grain foods and products; (4) choose healthy sources of protein (mostly plants; regular intake of fish and seafood; low-fat or fat-free dairy products; and if meat or poultry is desired, choose lean cuts and unprocessed forms); (5) use liquid plant oils rather than tropical oils and partially hydrogenated fats; (6) choose minimally processed foods instead of ultra-processed foods; (7) minimize the intake of beverages and foods with added sugars; (8) choose and prepare foods with little or no salt; (9) if you do not drink alcohol, do not start; if you choose to drink alcohol, limit intake

Fascinating. It appears they also use punch cards for their catalog photos: https://silverviscount.co.uk/image/cache/catalog/viscount-im...

Is that true? I have never seen this done for any image compression comparisons that I have seen (i.e. only data that is specific to the image that is being compressed is included, not standard tables that are always used by the algorithm like the quantisation tables used in JPG compression)

that's the point of the article... it's hard to write a regex that works on arbitrary emails from recruiters, but GPT is quite good at extracting information from arbitrary emails.

Well... I have deeply explored a scientific field (in computer science, not math) the way you are talking about. By reading other's research papers, implementing what they proposed, trying to improve upon it, coming up with novel solutions etc etc.

Very little of what I used on a day-to-day basis came from rote learning, a lot came from the deeper understanding that I got from actually actively engaging with the concepts.

I'm not proposing doing away with schooling completely, but the model should change (much in the way more progressive schools are doing anyway, just a shame that not many parents can afford to send their kids there).

and... where's the problem? That's exactly the kind of group problem-solving that is actually required in the real world.

Or we could accept that in this day and age, expecting kids to recite factoids broadcast to them by a "teacher" doesn't make any kind of sense any more, with google in everyone's pocket.

And testing in the typical style that fails when a student can use a textbook or google makes just as little sense.

The rumours are that it was too good at generating nudity for their comfort, and in particular that some users may have combined that with younger subjects.

The “hot new thing” is NeRF, neural radiance fields, which can take into account the way light interacts with the object (and hence you can correlate data from pictures taken at different angles)

Well… midjourney used stable diffusion (with an additional guidance model I believe, not just prompt engineering) for their beta model which they already closed down again… it’s back to their old far inferior model.

I think it will be more like pair programming. Already kind of feels like you’re pair programming with an intern with copilot.

Knowing whether or not it’s giving you a sensible response is one of the things that are hard for gpt-3, unfortunately. It has no concept of failing.

There are scenarios where a multi-party ledger can be useful in business... Think about a logistics chain where everybody needs to know that a record wasn't modified in retrospect.

But then you only need a mini-blockchain along the lines of what IBM calls blockchain, you may have a few industry players and say a customs agency agree to sign the entries and then there will be no dispute about whether an entry was signed on a particular day. Energy cost of that is negligible. (and you don't need to put the actual transaction info into the ledger... just the signature is enough so everyone on the chain can confirm later that it wasn't modified).

All of this is also pretty ancient tech by now... IBM just basically jumped on the blockchain branding bandwagon I'd say...

For sure. Hopefully the digital currencies will price themselves out of that market (and perhaps put pressure to put the less energy burning protocol variants into production).

To really give the right incentives to businesses probably requires some form of carbon taxes, very hard to get right but not sure we have a choice as a lifeform...

There's a lot cheaper ways to burn energy than bitcoin miners though... and arguably ones more beneficial to society. Water desalination for example. Bauxite smelting. Hydrogen production.

> Would Nike be happy if Portugal’s Cristiano Ronaldo suddenly appeared in Adidas shoes?

Well... it is happening the other way round, with runners wearing Nikes and painting over them to try and not annoy their other brand sponsors: https://www.businessinsider.com/nike-vaporfly-shoes-runners-...

But shouldn't that concern go the other way? If the Chinese government is dictating content policies of TikTok, that should obviously be a concern, but I have not seen this alleged.

Granted collecting say interest data based on viewing figures would be a precursor (I suppose it's likely that TikTok could figure out if a customer leaned left or right politically). But again that's not what the complaint is about.

It's also hard to ignore that TikTok's responses are at face value reasonable (such as evaluating touch patterns to identify bots... that's obviously something they have to do given the predominance of bots on social media platforms).

I actually think it's naive. Sure if military personnel or state department use TikTok that can create security risks, but the same is true with any other social media network.

The additional data that TikTok tries to gather from the mobile OSs is insignificant compared to what you can get from videos that people take willingly.

Russian soldiers use VK, that doesn't stop OSINT researchers tracking them down on there.

Personally I feel this is as much a failing of of mobile OSs as anything else. As long as we have an environment where ad networks think it's desirable to collect this kind of information it's going to get collected, whether by TikTok or ad networks. Apple at least attempt to limit what data is accessible by apps, Google is less incentivized to do so.

It's not much help to "ban TikTok" while the ad networks run wild. If we're going to be worried about China I doubt it'd be too much trouble for them to place a few engineers into ad networks and exfiltrate the data from there.

In defence of 'security' screws, they are commonly used in applications where it is a regulatory requirement to make it impossible to open an enclosure with common tools, such as power bricks which have exposed mains wiring inside. They are a great compromise here, as they can satisfy the certification authorities while still being accessible to a motivated end user with an amazon driver kit.

The point (in theory) is that in efficient markets you can choose between different companies offering a product/service and that competition drives down prices / costs. The healthcare market in the US is very much not a competitive market (in part due to transparency issues which this is trying to address, in part because of the perverse incentives in the health insurance system).

It's not so easy to have a competitive marketplace for government services (though I would not be surprised if the US saw the emergence of private police forces now, for those parents who actually want police officers that have a duty to protect their kids...)

Just be aware that that's a whole lot of conjecture with little solid evidence behind it. Saying "these brain injuries historically resolved but don't any more" sounds much more like a change in diagnostics to me.

The evidence against seed oils / processed vegetable oils is not very compelling yet (here's a pretty balanced summary: https://www.healthline.com/nutrition/are-vegetable-and-seed-... )

They definitely got that product placement for free, and the license to the movie clips for their advertisement. Wait...

Indeed, I do make an effort to write if(ptr!=NULL) (partially because it more explicitly encapsulates the intent, much like writing ptr=NULL when "clearing" the ptr) but changing the developer expectations around this sounds like a terrible idea.

I don’t think it’s quite as simple as “take people/algorithms at their word”. “Sentience”/“Consciousness” are at least in part social constructs, in that the words we use, both between humans but also in our internal dialogue, are social constructs.

I think it is hard to untangle any potential non-linguistic aspect of these concepts from the language we use to describe and perceive them.

Currently we have the challenge that algorithms can regurgitate the language we use to discuss these concepts, and we don’t have the tools to distinguish that from how humans use those same words.

With current transformer models we can make a cogent case that there is a fundamental difference to human mental states, however I believe this will very soon get a lot more difficult once multi-modal algorithms are used in embodied forms and with continuous re-training and context databases.

You might find "the Ordinary" brand from Deciem appealing (or Niod for more exotic ingredients), simple products with ingredient concentration specified.

Ah, that’s fair, I thought you were arguing that the health benefits seen from weight loss can be attributed to healthier diets alone.

This made me wonder whether potatoes are really this expensive in Australia? In the UK buying 2000 calories worth of potatoes from a major supermarket would cost the equivalent of AUD2.12 not AUD9.90 ...

Though by comparison rice gets you 2000 calories for AUD1.03, spaghetti AUD0.92, carrots AUD2.51, "pure vegetable oil" AUD0.059 so a more varied diet could be cheaper.

Actually I'd like to propose the carrot-only diet... Trying to eat 5kg of carrots a day is bound to make you lose weight, though your skin may turn orange ;)