ntonozzi

He’s not saying that everyone had lower productivity through the pandemic, but that those hired during the pandemic and working remotely had much lower productivity.

Don't forget cars.

Check out https://github.com/erikbern/ann-benchmarks for some benchmarks on some of the different ANN libraries out there. I'd be interested in hearing other's experiences using these libraries in production.

Typically and hopefully they would have to return the money that they did not spend, and it's better for the investors to get that portion of their money back than to have founders run it into the ground.

Perhaps more carbon efficient, but 15x less energy efficient:

> Biking takes around 25 kcal/km [iii] above basal metabolism, which is equivalent to .11 MJ/km. A typical car in the US gets 25 mpg, or 9.5L/100 km, which is equivalent to 3.3 MJ/km. The Toyota Prius takes only 5 L/100km, or 1.7 MJ/km. So a typical car takes 30x more energy per kilometer than biking, and a Prius takes 15x more. This is what we expect given how much heavier cars are than bikes.

Llama is trained with _more_ data than is chinchilla optimal in order to make it better and cheaper at inference time, instead of just getting the highest quality of model that you can based on a given training budget. Llama has fewer parameters and was trained on more data specifically so that it would get high quality results on cheaper hardware and be easier and faster to run at inference time.

If you only deposited $1000 they are not going to give you $250k back.

Designers sure love gradients+noise. And dang does it look good.

My suspicion is that this is true for you, but untrue generally. When walking and biking, all of my close calls have been with large trucks. Drivers typically feel more confident going faster in large vehicles because they are higher up, and they are less likely to see pedestrians and especially small children. Modern trucks and SUVs are very unlikely to rollover in normal intersections, so the more awkward turning doesn't force drivers to slow down. Is there any data on how often trucks and SUVs hit pedestrians and bikers compared to cars? They are certainly more lethal when they are involved in a collision.

How are they different from models that already exist like FLAN-T5?

Including the word 'phone' six times in a popular blog post is not going to help their predicament.

I think Flan-T5 is fast enough, but I don't think it generates text or abstract reasoning at nearly the same level as current GPT-3 models. This indicates a deficiency in the benchmarks and metrics that we use to evaluate LLMs. For generating embeddings it might work well enough though.

For the most part I agree, but Storj seems like it is at least attempting to solve a problem not created by blockchain: https://www.storj.io.

Amazon emplyees actually have a negative cash salary that they pay back with equity they sell.

Why do you say logistically no? The challenges of distributing the runtime?

Java, Go and C# (and node) have very similar performance, e.g. https://benchmarksgame-team.pages.debian.net/benchmarksgame/.... For all of them, the key to writing high performance code is avoiding allocations and boxing. Go and C# both do this slightly better than Java, but in most domains where these languages are used, this is not a big difference (and this is where you might use C/C++/Rust instead). I've found Go to be more verbose than Java, but I haven't used Go much since generics were released.

Go belongs in the exact same bucket as Java and C#.

I agree MacOS feels dated, some examples: The Windows window manager is far better than the MacOS equivalent, the UI is higher quality, more consistent and more discoverable, and PowerShell is better than zsh.

Oops, good point! EFF also has some good articles about software patents: https://www.eff.org/issues/patents.

Can you explain why it's not what EFF means? To me, the part of software patents that doesn't make sense is that you can patent any possible combination of ideas — in this case the combination of using voice and remotely controlling a computer — and it's considered a valid patent.

In my opinion, this is like patenting putting a blue Lego brick on a red brick because nobody bothered to write down or patent that particular combination before. There's no core novelty here, and obviously Mycroft was not stealing profits from Voice Tech Corp.

Software patents are a huge drag on small companies and innovative large ones, it's just that large companies are better able to handle the costs. Our world would be better if we ended software patents. Check out FSF's campaign to end software patents: https://endsoftwarepatents.org.

What did you use as the index structure for the nearest neighbor search? Did you use a library like FAISS or Annoy?

Was it PRQL? That’s what all of these comments remind me of.

https://prql-lang.org/

I see, thank you for the insight. The OpenAI paper about summarization is an easy read and very similar to your approach, if you are looking for somewhere to start: https://openai.com/blog/summarizing-books/.

Have you compared GPT with prompting to other models like https://huggingface.co/sshleifer/distilbart-cnn-12-6? Do you know what the state of the art is in summarization technology?

You can get a CGM from Levels Health (https://www.levelshealth.com/) but it will cost you and I don't think it will be covered by insurance.

And yet many job postings, even those that are only in Colorado, still don't have the salary on the job posting.

.

Some extra context: https://en.wikipedia.org/wiki/Fall_of_the_Western_Roman_Empi....

No, there are plenty of ways to violate a boundary besides rape.