gala8y

I will switch soon, just need to collect all stuff and find time. Waited a bit for nvidia drivers to catch on as I want some models to run locally.

Thanks for the talk. I tried to learn how bad X was in terms of security and it wasn't easy to find anything concrete. Took me a while to realize it was a mess.

Look up "macropads" on the Internetz.

> Is there anyone currently capable of developing an AI soulmate

I believe there is. Here is a video from yesterday about Chai, which supposedly has 10M active users [0]. Title changed from 'Chatbot platforms might replace social media' to 'Chatbots are becoming virtual reality [Sponsored]' in the meantime. The video is about how they experiment with models and metrics rather than actual user experience, but I guess they deliver on 'capable of developing an AI soulmate' front.

[0] https://www.youtube.com/watch?v=ILZ45sxon-4

> Very "Old Web" vibes.

*Owl Web

Owl Rights Reserved (at the footer)

Yesterday I tried to vibe install (TM) marker-api docker image and failed miserably. Still, I was able to try. :)

I would conceptualize it rather differently. Let say you have two daemons inside, Fast Coder and Detail Scrutinizer. It is about checking how much power every one of them is given on two separate scales (every each of them working in their dominion). If you switch on only one and neglect the other, you can get into trouble (or not, depending on the circumstances). Anyway, these are two variables, two facets, and knowing how to mix them appropriately makes the whole difference. It's like internal combustion engine mixing few kinds of fuel. What you get at given moment depends on the mixture of these different qualities.

Cf. "A real-time 3D digital map of Tokyo's public transport system", https://news.ycombinator.com/item?id=37829061, Oct 2023

Nice play with the name (Timefold).

Just a quick, too-simple thought... if someone searches for shoes online or anything else to buy, GOOG will deliver and make a buck. All other scenarios... they don't care.

Mind you, awakening does not equal not awakening.

If I have a friend doing... stuff, he/she already has my support and they kinda know it, you know. This is superficial, 7 yrs old kid's approach. Plus, as others mentioned already - it's none of your business to meddle around with their or mine contact data. No, thanks. Real communication with real people for me.

Loved del.icio.us. This is a pet peeve of mine, that proper tagging (with AND & filtering properly implemented) is rare in software. Tagging peaked with del.icio.us.

> A failure to do so means you serve some other end, perhaps one you do not realize you are serving.

What do you mean here?

Not bad, this gets rather steep!

Fiasco by S. Lem, work of fiction of course, is a nice food for thought here.

This resonates funnily with "I hacked this together last night for fun" a bit down the page.

Are you serious? Can you give some examples?

I will piggyback on your comment. Absolutely stellar example of fine grain thinking which accidentally shows huge expertise and real life experience of the author. Regardless of type of a problem, the amount of thinking put into solving it makes the difference. It is thinking in systems, building complex mental models and finding edge cases which pays off, regardless of the domain / problem at hand. People tend to halt too soon in building mental models. When you are responsible for any given area, you better build a fine grained model. Obviously this is costly in terms of time, money, lost opportunity,... and there is of course a blurry line where you should stop building complexity of a model and just implement a solution. Life will come knocking on your door anyways, Aunt Entropy shows up sooner or later.

This is also why almost all news is non-sense for an expert in given domain. Basically... "It's not that simple."

Oh... why... oh why... I will never understand such behavior. What are incentives here?

Unfortunately universal solutions are often not enough. Good thing is for non-programmers to code a little, script a little.

Good luck with tuning your PS a bit. It should pay off somehow, me thinks.

> It's an interesting question (how much density is "right" for productivity apps).

Between extremes of GUI presenting all functions at once and accessing all functions through memorized keyboard shortcuts, there must lay large lands of possible, experimental GUIs which unfortunately never get tested in popular software (which defaults to sub-menus + some keyboard access). I mean zoomable UIs, radial menus, 3d concepts in UIs,... - it's hard to see any of research on experimental interfaces make way into actual everyday-use apps.

Every standard GUI element could be thought over. Once I worked with some standard accounting software, where I needed to select some stuff many times during the day from a dropdown. I quickly learnt how to move content from a dropdown menus into a spreadsheet with some formulas which allowed me to paste selection I needed back into these dropdowns 10 times faster. When I was leaving this temp gig I showed this solution to some guy in accounting, whose life was basically opening these dropdowns all day long. Regardless that he was really mean guy, I showed him the trick. He almost had tears in his eyes seeing it. I am not counting muscles strain avoided... Multiply by thousands of users everyday...

Not to mention trying to integrate interoperability of apps within the OS (e.g. to have one program to check for spelling in all apps, across the OS - I cant remember name of one such fine try (ages before)).

As per PS, have you ever tried to really personalize PS UI to your liking? I remember friends getting really opinionated on positioning all those menus perfectly for themselves (and for a reason). Personally I feel like PS is very intuitive for me, whereas in similar apps I cant do simple things, but it's probably because I was kindly shown around.

This complexity also shows in 3d software. I love all these UIs like 3dmax, SoftImage (!), Blender, Lightwave. Creating 3d has so many aspects to it and it shows in complexity of these UIs. Generally speaking, I much prefer being able to click one of many things shown on the screen than clicking through endless sub-menus, which was always a waste of time for me.

There is also Sayboard (open-source, multiple languages): https://github.com/ElishaAz/Sayboard

Thx. Nevertheless, https://xkcd.com/1053/ still applies. I'm one of lucky 10000 today.

There is a Hacker News Comment Highlighter add-on both for FF and Chrome based browsers. Highlights what is new after a refresh.

Revival was on a soundtrack of a (mediocre) movie Reality Bites. It has wonderful sense of South-style communion for me. I remember that was the only song by Me Phi Me, which I enjoyed at that time.

Thx for links, good stuff. Let us keep it up.

> Training data is the competitive edge, so everyone refuses to elaborate about what they use.

I absolutely understand it, thus curiosity about some leaked insiders knowledge or anything. Anyway, given availability of these huge data sets (incl. ones you mention) and advancements in processing techniques, we probably have a long way to go. Also, video looks like two guys talking about everything except AI bubble.

How much of existing data did they devour?

- internet (incl. blogosphere, all www, forums, social networks, dark net,...)

- all of digitized books

- all statistical data, governments data, military data

- all scientific papers and books

- all commercial companies data

- all private data/personal data

- ADDED - all video/audio data

- all new data...

If someone has some links on what the current state of affairs is (what they really trained on), I would be really interested.