comboy

But it's still a mobile phone transmission power (1W?) and then r^2. Distinguishing that from noise seems mind-boggling.

Do they have anything in common with Mandarin Blueprint? It seems to use very similar mnemonics. How far did the tool get you?

There are apps now that provide you with listening material improving your comprehension while also doing some SRS on it. Superchinese focuses on pronunciation a lot, Clozemaster on listening. HelloChinese and ChineseSkill are more duolingo-like but way better for Mandarin.

And Untamed ("Leaving soon") is still on Netflix, it may seem cheesy but out of things I tried watching it has surprisingly clean, easy to understand language with lots of common phrases while also offering some interesting story.

That's my point.

So there's this short little book "Food rules" by Michael Pollan. Not much content but seems like the author went through a lot of research. He comes to conclusion based on this tons of data that all we really know for sure is that people living in these blue regions are living much longer and it seems to be related to what they eat. That it is basically the only solid and stable data point we have. Welp. (I'm overstating it a bit, but not by that much)

Generative AI is much easier than photoshop the same way that photoshop was much easier than trying to alter photos any other way. In both cases it is a completely new level.

It's ban photoshop all over again.

It's just the new reality. Porn is not even the worst thing. If anything it can make real photos/videos harder to get viral.

But augmentation will be (if not already is) used to make very tiny adjustments, like you make one candidate/celebrity face slightly more symmetrical and the other one less, change their gait just a little bit. These changes are non-perceptible, are not interesting enough to make a scandal, and they can have huge impact on how people perceive other people. I mean, I am not suggesting that it isn't a puppet show without those, but the game is evolving.

Groundwater mostly, but also clouds (minuscule amount).

Even within tech circles surprisingly few people are aware how ubiquitous power law distributions are.

This makes sense for idle devices, but say airpods know they have been opened (or other bluetooth headset can be actively listening for a few seconds after power up), and on the other device you explicitely click connect on already discovered divice. I also don't understand why these connections are not 1ms even for devices which were not paired previously.

So what do you think about the paper?

I love how well this is written. Definitely "look how interesting this is" rather than "look how much do I know". And it dives as deep as needs to, while being accessible for almost everyone. One really needs to master some topic to be able to describe it simply. Great job.

How about the gov comes to you and tells you that you must keep all deleted photos on the device and that you must not tell anyone about this request.

I'm surprised at such a mean comment and lots of follow-ups with agreement. I don't know Sam personally, I've only heard him here and there online from before OpenAI days and all I got was a good impression. He seems smart and pretty humble. Apart from all openai drama which I don't know enough to have an opinion, past-openai he also seems to be talking with sense.

Since so many people took time to put him down there here can anybody provide some explanation to me? Preferably not just about how closed openai is, but specifically about Sam. He is in a pretty powerful position and maybe I'm missing some info.

> LLMs are state-of-the-art at detecting sarcasm.

This is such a precious gem.

Full quote for those unfamiliar because it's so good:

"All that you touch, You Change.

All that you Change, Changes you.

The only lasting truth is Change.

God is Change."

What if you didn't know? You wouldn't know whether you like it until you've learned more about the artist?

That's a beginner mistake, it's really hard to make if you dealt even a bit with some regular expressions.

Plutonium half life is like 90 years. It wouldn't get to the top 10,000 list of bad things we did in the last 100 years.

thanks

Many commenters seem to be using magic instead of file, any reasons?

It used to respect custom instructions soon after GPT4 came out. I have instruction that it should always include [reasoning] part which is meant not to be read by the user. It improved quality of the output and gave some additional interesting information. It never does it know even though I never changed my custom instructions. It even faded away slowly along the updates.

In general I would be much more happy user if it haven't been working so well at one point before they heavily nerfed it. It used to be possible ta have a meaningful conversation on some topic. Now it's just super eloquent GPT2.

I love it, thanks :)

Outstanding blog post. NRF24 had been my fav for a long time, I planned switching to LoRa but I didn't know the latency is that high.

Sure, touchscreen is expensive, but if you just have notification on your phone that the door is not closed properly that's already considered "smart". Basically anything connecting to your home network is advertised as such.

I wouldn't. Cost to make a device "smart" is much less than $1 in parts, app and infrastructure are not rocket science and even with no evil intentions you get very valuable data on usage, what breaks first, environment, location etc. Seems like a no-brainer.

Those will soon also not require you to connect them to your wifi, NB-IoT both chips and plans cost for businesses are already very good.

"Thanks to our best sensors yet, Apple Car will not run over any person wearing any Apple product" (applause)

What use cases? I had my cornea scanned with it when it was damaged some years ago. I didn't appreciate that it had to be touching it.

I said rhymes. You are presenting argument against something which I did not write.

It's worth remembering though that it all drops down to trust anyway. What source do you consider trustworthy? Journalists write false stuff all the time (you may know some specific people you can trust), so that leaves research papers. Reliability of data there is higher than from a random sample on the web, but it is very far from something I would trust blindly.

Any human compiling data from different sources can also make a mistake. Can fail to use reason when it should be clear that the data is not consistent. I don't think there's any doubt that in some time, such error rate will be lower for large models than it is for humans.

s/used GPT/used Google

history rhymes