donkeyboy

I had no ifea they were struggling. Tldr; their competitor Insta360 is battling them, and they have YoY revenue drop.

Gopro has this cool reliable aura around them. How could they he struggling? So bizarre

Data.table is just so much faster, and the sql-like stntax is easier tonunderstand

Thats what came to mind. Daft Punk, the Alive 2007 recording

Sadly this doesnt have a simple way to know how to win besides the ForceEven approach and offering an Anki Deck. I wish there was some more intuition or guidance around winning that humans can memorize

Can you explain this to me?

I think the main point is that these two companies have been thought of as equals. Yet OpenAI has a 850B valuation while Anthropic has a 380B valuation. So that means either OAI is overvalued, or Anthropic is undervalued. If you believe they are roughly equal, then it makes sense to have demand for Anthropic.

Cute story. This reminded me how in elementary school and middle school I used to draw pencil drawings of rollercoasters on my page to pass the time. Rollercoaster tycoon fan :)

My bad

Its missing the funniest signature. Back when Spain still had monarchy, King Ferdinand VII simply signed his signature “Yo el Rey” - literally “I, the king”.

This is amazing.

Elasticsearch / Opensearch is the industry standard for this

Cool idea. One thing the author might like: if you sign up to lastfm, it tracks all of your music plays. So you can go through the data and see what you were listening yo each month/week

Want to read part 2!

The documentation for common ai packages is pretty good too. For example, pytorch docs, peft docs, timm docs.

Thank you, the future is awesome!

Finetuning is pretty much necessary for regression tasks. Also useful for classification since you can get the direct probabilities in case you want to do some thresholding.

From what i read, this post doesnt announce we’ve found some crazy extremophile unicellular microbe. Just that there is evidence to suggest they are there (due to the chemical makeup of soil/boreholes).

Imagine it’s raining, or they come really fast. Even if not so, it is always expected to tip the person doing delivery. That’s just the custom, like tipping in restaurant or tipping the bartender is the custom.

A very wholesome read. Thank you for sharing. I’ve never been so into outdoors/camping/fishing, but it made me reflect on some of my adventure trips I’m doing right now while I’m still young. And maybe these will be talked about in my future family.

I agree. And i think other comments dont understand how utterly difficult this is. I think that there is a translation tool underneath that translates into English. I wonder if it can also figure out binary ascii or rot13 text. Hex to letter would be a very funky translation tool to have

Concat could work too although less efficient because you need to make a new tensor.

Actually summing might learn a concat on its own. Imagine the embedding learned for a token takes up the first N-20 dimensions and leaves the last 20 dimensions as 0. And the positional encoding causes the first N-20 dims to be 0 and the last 20 to encode the information. Then when you sum you are actually concatenating. So I think of them as equivalent except add is more efficient/preserves the dim space, while concat would grow the dim space. And for something like position, which certainly does not need to occupy 1000+ dimensions, it would not make sense to concat all of that since it would be wasteful

Looks like xyzzy and plugh originated as a magic word in the computer game Colossal Cave Adventure

Those things look freaky! And the legs seem like they function more as whiskers/feelers. I wonder if they are swimming from the wings or if the “legs” are helping push them forward. If the latter

This video is way too funny. Thank you for sharing.

Plastic is made from upstream oil and gas companies. Kind of hard to be net-zero in that sector

Different topic, but can someone talk about which browser they use now? I’ve used Chrome for the past N years with ublock Origin, but recently I’ve been getting some ads on Youtube. Also some websites just didn’t work in Chrome. I switched to Brave a week ago and things seem ok, but it’s a weird browser with Tor built in, and also Spotify.com always crashes with a Memory problem. Does anyone have any thoughts on browser preference these days?

Cool to see this idea in 1996. I think the main problem is that you have to write up / document everything for the Remembrance Agent to retrieve. The large majority of human knowledge stays locked to the brain, and sometimes our attempts at converting brain knowledge onto paper (“writing”) will leave omissions.

There is a paper on arxiv saying that GPT4 correlation with human evaluators on a variety of tasks with strongly positive. I am also uncomfortable with it, but using GPT4 as a grader is not as bad as you think.

I think this could be fake. He says it’s an MoE model, but then explains that it’s actually a blended ensemble. Anyone else have thoughts on that?

Actually, AIs have been training on AI-generated content for a long time now. An example of this is student-teacher distillation.