As an aside, if you actually want to ask the wire cutter random questions, you can check out the green bubble in the lower left here: https://askthis.site/nytimes.com/wirecutter
I've changed my mind. Historically I've been opposed to computers being a panacea for education.
At the MIT Media Lab I was one of the earliest critics of the strategy and effects of the One Laptop Per Child project.
I've changed my mind.
There are real issues certainly, beyond AI making things up, encouraging risky behavior, or kids asking it naughty questions. Things like, can it damage a kids social skills, create echo chambers, or push them towards propaganda.
But the upside is incredible. Integrating an LLM like GPT4 into a more kid-centered suite of models unlocks tons of potential.
The first thing I built into my AI tutor system was the ability for it to look up facts from reputable websites, and to help kids answer questions for themselves.
Then I taught the AI to help a kid develop their interests, so they take a topic they're curious about and dive deeper in to it.
And then I taught the AI to notify me if there are topics that are more appropriate for a parent to engage in.
The tool I'm building actually helps kids become more curious, creative, and thoughtful.
It isn't perfect yet, but the truth is kids spend 7 hours a day on screens already. I'm very excited at the possibilities.
I've built a tutor for my kids, that listens to them, and helps them explore topics that they're passionate about.
What's blown my mind is how it has driven creativity OFF screen too. They engage with the world in a deeper way when they have the power to explore their passions on their own.
If you want to try it, join the waitlist here: gotwonder.com and then email me: seth@gotwonder.com and I'll bump you up the list!
I was working on this yesterday. It seems that the most common approach with Whisper is simply to break the audio into chunks and transcribe each one separately. This works but as you'd expect sometimes has trouble at the edges. The segments also have to be sufficiently long (like 10s) or the accuracy suffers, meaning it's not truly real-time.
You could do better by overlapping the segments, except then stitching the transcriptions together becomes an issue since whisper doesn't provide reliable per-token timestamps [0], and the output of the common part of overlapping segments isn't necessarily the same. I can imagine a cool approach where you transcribe long, overlapping chunks in real-time and intelligently merge the stream of words somehow though.
Some more useful discussion here (whisper.cpp project, but still relevant) [1].
It definnitely takes some set up to make it work, but once working, it is pretty reliable. It requires an occasional merge conflict resolution, and setting the right files to be ignored.
I love that it is just folder with markdown files.
One of my favorite versions of this trick was invented by Martin Gardner and actually uses a timing attack... on your brain! [0] Without apparently telling the magician anything, you end up giving enough information to uniquely identify your number!
Very fun!
If you wanted to add one more level of misdirection, the first cards could be textarea that you have to select with your mouse and hit ctrl-c to copy ;-)
This would hide a bit of the method for the twitter reveal.
-Seth
(I also make online magic tricks magicseth.com/animal )
And this may not be the effect you are going for, but you could make the final twitter link go to a specific tweet, where you say "I was hoping you would copy the 3 of diamonds!" and it would be right :-)
I'm pretty sure mine were some of the criticism you heard ;-) My critique largely focused on the imperialist attitude of the project: Negroponte described the project with hope that it would be as impactful as the "Niña, the Pinta, and the Santa Maria" without any sense of recognition of th negative impact of those ships. When the first technical bring up touted OLPC developers running Doom on it, it was a reminder of how, no matter how much we claim technology is absent of agenda, it is many ways a reflection of the values of the developers.
Everything done by anyone is a reflection of them, their needs, constrains, ideas, and knowledge. Intention matters, and inaction is intention-less.
However I don’t think your metaphor is apt — OLPC didn’t intend to be a conquest in any way that was about extracting resources let alone bring resources those back home.
The OLPC didn’t replace something, it created something. It pointed a way.
Does it really? We all know the saying about the road to hell, and I suspect the people on the receiving end didn't particularly care about intentions, especially if the result was a net negative.
I seem to remember Carlos Fuentes narrating an entire series "The Buried Mirror" starting out with that (he immediately follows it with "[Columbus put his newly-discovered] noble savage in chains") but I don't see anyone calling him racist.
