mFixman

This is the case in the UK (with 60 days instead of 90), and I think it's also the same in the US.

You can still take information from the hexagons that are different from their surrounding ones.

I know exactly the intersection where that dark part of Hammersmith Road is.

Bromley is actually a fantastic borough for cycling: mostly rural, great walking/cycling trails, good rail connections for the way back, and it even has weird Victorian era dinosaurs in Crystal Palace Park.

Why cycling's not more popular there is a mystery to me.

This is not about people, this is about journalists.

Almost everybody who can use a LLM understands its results shouldn't be taken literally. Still, it takes only a couple of clickbaity "parents concerned about ChatGPT turning their children into nazis" stories to move public opinion against LLMs.

There are a lot of serious open legal questions against these models, so preventing these stories is on the best interests of Microsoft.

If ChatGPT had existed in 2016, the media would have blamed it for Trump getting elected.

This is fantastic.

There are very few authors that could accurately predict challenges with future technology, but Asimov nearly nailed it with the workarounds around the three laws of robotics.

I don't understand. How would a marketing agency plant a job posting in the Microsoft careers site?

The same argument I make against smartwatches: they are great for people who regularly wear the dumb version, but uncomfortable for people who don't.

Smartphones became ubiquous quickly because by the year 2007 almost everyone was used to having a cellphone in their pocket, smart or not. Personally, I find regular sun/glasses uncomfortable and I don't see myself ever buying smart ones.

It seems like a complicated solution to a reasonable problem.

Text editing involves a pop-up keyboard. Why not add a menu with arrow keys, selection toggles, and copy/cut/paste there?

> Birbs stagger as they walk: they are reduced in alternating associative order, starting with left associativity at birb index ⌊len/2⌋:

    BB -> (BB)
    BBB -> ((BB)B)
    BBBB -> (B((BB)B))
    BBBBB -> ((B((BB)B))B)
    BBBBBB -> (B((B((BB)B))B))
    BBBBBBB -> ((B((B((BB)B))B))B)
   ...

This is the worst operations order I've ever seen and I love it so much.

My main pet peeve is how you cannot hide major roads from Google Maps, but there's no way to show trains and most public transit without clicking on a station. Even on a large city like London half the public transit is invisible.

What alternatives are there? I'm trying to use OSM, but I find it to complicated and time-consuming to do anything.

That video is so bizarre. Prigozhin is going way beyond the "2014 status-quo was good to protect Donbass Russian-speakers but 2022 invasion was a mistake" story that's common in Russian anti-war campaigners and actually denouncing the last decade of Russian military incursions, of which he was a major player.

What is he even trying to do?

There's no chance of an actual coup without support from the people or the army, and he's hated by both. His usual dramatic requests for more money and equipment won't work if the first thing you do is denounce the War, and nobody with morality or a conscience would have become the head of a Russian paramilitary group.

I'm as confused as anyone reading this. Is all of this an extremely risky bet to get strategic control of the Russian army? In that case it seems to have failed spectacularly.

> In particular, 妛 was probably created when printers tried to create 𡚴 by cutting and pasting 山 and 女 together.

It's crazy to think that a lot of written Kanji and Hanzi come from print drifting and a small accumulation of errors, similar to how many new words are formed in spoken language.

Is that like a LLM Stable Diffusion? Neat.

I don't have a PC with a powerful GPU. What's the easiest way I can play with Llama on AWS, Google Cloud, or somebody else's computer?

Trees and crabs and birds appeared very recently in the timeline of life on Earth and will probably get extinct eons before shrimp or tardigrades know what happened.

Life is likely plentiful in the Universe. Earth-like life is almost certainly unique.

The possible shapes of life are much bigger than the universe.

The chances of intelligent life existing outside Earth are lower than the chance of that same intelligent life speaking flawless Swedish. We might as well send some Europop records to space as signals to aliens.

"Intelligence" is a human invention created by humans to describe humanity. It only applies to humans and human-like beings.

Meta gave everyone free Oculuses and encouraged Metaverse meetings back when I worked there.

Almost nobody used it because the product required a lot of setup, was downright ugly, and was completely inferior to the usual Zoom / Workplace Chat meetings in every way.

The whole thing had a very top-down feeling, with directors trying to convince managers to make their teams use the Metaverse while nobody in the company really knew what the Metaverse even was. It's great that Meta was still mildly engineer-driven and didn't force employees to use it, because in that case they would have ended with a Microsoft Teams situation.

Computers are becoming smaller and the internet is becoming even more ubiquous. I don't see your future happening.

Why would I spend a lot of money, space, and maintenance time for a computer with a good GPU when I can just use any of the two internet connections I'm already paying for? I'm already trusting corporations to read my email, and no ChatGPT answer can get more personal than that.

This is also why I think cloud gaming is only going to become more popular.

Can't you store the database locally and then have Dropbox or Google Drive upload it to your drive?

It's still not 100% secure, but you would need both a compromised Keepass app and a compromised Dropbox app.

From my experiences with Google integrated AI, I'm skeptical that Google can win any battle with it.

ChatGPT and other modern AI tools are useful specifically because they do what I want when I want to do it. Compare this to Gmail's smart compose: by the time the AI makes a suggestion I'm already half-way through typing that text, so its prediction is completely useless even if it's correct.

(Also, I really enjoy the random historical rambling at the end of some Stratechery articles).

Adversarial machine learning is an extremely hard problems space.

If you have a single static model and your adversaries can react to it, then it's an almost impossible fight unless you are willing to give a ton of false positives and block a lot of perfectly valid prompts.

Microsoft cannot train a new model every second, but attackers can change their strategy depending on the Chatbot's answers; any "Is this user trying to access your prompt?" would be broken easily.

You could say the same thing about any >$200 phone, or any new device at all: gadgets are luxury devices, and there's no upper price for the luxury sector.

Besides, if this lasts at least 4 years then the amortised price is more reasonable.

> 0 and -0 being distinct values is honestly just an accident of convenience with how floats are stored

This is actually a valid distinction because these values can represent different values.

Floating point values are not numbers, but ranges of possible values: float -0 is a number in the range (-FLT_TRUE_MIN / 2, 0] and float +0 is a number in the range [0, FLT_TRUE_MIN).

IMO if I were to redesign IEEE 754 I would add a third float 0 value that always represents exactly 0 and redefine float -0 and float +0 as open limits, but that would probably make everything a lot more complicated.

For the vast majority of human history, the only transmission of information between generations were oral stories. Modernity defines nations by their borders, medievality defined it by their feudal alliegances, but true antiquity defined it by their mythologies.

I always wondered how much was lost in the long transition from the natural oral to the new written tradition. Do we humans learn things differently when hearing them than when reading them? After all, voice is multidimensional: pronounciation, tone, gestures are completely lost when conveyed into text. Why do people go to political rallies or speeches instead of just reading their contents? Why do children go to school instead of just reading books?

History is full of lettered peoples being guided by oral traditions. Maybe there's a good reason why the Rabbis took 1500 years to write down the other half of the Torah.

I would put something between the plough and the full mechanisation of agriculture.

You can separate eras in human history by the the dramatic jumps in the amount of people that can be fed by a set amount of human labour.

> His immediate concern is that the internet will be flooded with false photos, videos and text, and the average person will “not be able to know what is true anymore.”

Isn't this the case already? I expect every post I see in large social media sites posted by somebody I don't personally know to be non-organic feedback by a social media expert.

People are doomsaying over a scenario that's identical to the present world.

Also fantastic for those of us who only have a work laptop and don't want to buy an expensive console or PC to play the occasional video game.

If I read that correctly, does it mean that Epic can now direct you to a website to buy in-game currency outside of the reach of Apple's 30% bullshit tax?

That sounds like a major win against Apple.

One of the biggest liberties DeepMind had was a completely separate hiring process and pipeline than Google.

Their open positions mysteriously disappeared on November last year and they are still closed outside of specific senior roles and a very open ended "register your interest if you have a PhD".

Big loss for DeepMind if the separate pipeline is lost. Being able to hire for their priorities instead of whatever Google's hiring for is one of the reasons it was so successful.

They had Workplace, which was a fantastic product that got shelved externally when they couldn't get a big payout early in the project.

It was the perfect Slack replacement for companies with over ~1000 employees, and I wish my current company was using it.