vimax

5 ms·

91.
by vimax 3y ago
Yes, we should develop a new theory of language based on an information theory approach rather than a discrete symbol approach, or maybe something involving both. But you can’t say information theory as it stands gives us a completely satis
92.
by vimax 3y ago
Hard to apply those to software engineering as well.
93.
by vimax 3y ago
That’s best when possible, but if each context and response very large, a good prompt is the best way to distill the information without fine tuning. I suppose you could try asking for shorter examples to provide for a few-shot prompt, coul
94.
by vimax 3y ago
The Chomsky hierarchy will remain unchanged and very useful for constructed languages. I didn't mean break out, as in create a higher level. LLMs are not going to magically become more than a turing machine or anything. I meant we n
95.
by vimax 3y ago
To me it shows how fundamental language is to intelligence. Not just that LLMs are very capable by themselves, but significantly improve speech and image models. We’ll need to break out of the Chomsky hierarchies and develop some new theori
96.
by vimax 3y ago
Or just provide it with sample contexts and expected responses, and ask for a prompt to produce those responses from the given context.
97.
by vimax 3y ago
Thank you for this! I've been thinking about this since the digital artists got up in arms about AIs coming for their corporate jobs of shoving more ads down our throats. People with lots of money have always been able to buy other pe
98.
by vimax 3y ago
How about slipping a bank teller a note: Help, the man with me has a gun. He thinks this note is a robbery demand.
99.
by vimax 3y ago
Sure, but that’s not too different from the current scams where there’s some excuse you can’t talk right now. In both cases it can be solved by checking if it is coming from a known number and calling the person in question to confirm. Peo
100.
by vimax 3y ago
Why does it matter if you’re talking to an AI or a human? What can a criminal AI say that a criminal human cannot?
101.
by vimax 3y ago
Thank you, I’ll need to look into it more. The report is about the use of user number fraud and illegal transactions to enable key early insiders to cash out and leave the company in decline. That said, it looks like most of the accusations
102.
by vimax 3y ago
I’ll never understand why people are so quick to provide such personal data for trivial reasons.
103.
by vimax 3y ago
When the person stabbed was the CTO of a company just accused a week earlier of fraud and facilitating illegal transactions, it may not be that silly. https://hindenburgresearch.com/block/
104.
by vimax 3y ago
I find it strange that nobody is mentioning the Hindenburg Research report against Square that came out a week before the stabbing, and the fact that there’s likely a fraud investigation underway. https://hindenburgresearch.com&#
105.
by vimax 3y ago
It’s not very different from asking every question you have on stack overflow
106.
by vimax 3y ago
It’ll be a world where it’s important to know the right question to ask
107.
by vimax 3y ago
You can use an LLM to rewrite a document to a question answer form.
108.
by vimax 4y ago
Have you watched a video of someone with rabies exposed to water? I don’t recommend it, it is more than a hard time swallowing.
109.
by vimax 4y ago
I think the cycles are something of a distraction. Ultimately the problem comes down to this. Given a number N with prime factors (p1, p2, …pn) And N+1 with prime factors (q1, q2, …qm) What is the relation between (p1, p2, …pn) and (q1, q2,
110.
by vimax 4y ago
I don’t think (b) will work. The Dallas police pension fund is currently $3b in the red, how will they pay anything?
111.
by vimax 4y ago
I feel like you’re thinking about this wrong. Knowledge graphs are incredibly useful in combining language and structure, but often not worth the expense to build. Future language models will build knowledge graphs for us with little expens
112.
by vimax 4y ago
Completely unreadable. Who in there right mind makes a code block overflow hidden? Not sure if the blame is with GitHub or OpenAI, but how do either of them not open this on a phone at least once?
113.
by vimax 4y ago
This is a dangerous proposal, it includes adding a new HTML element attribute “anchor” that will completely reposition the element. I would expect there are frameworks out there using “anchor” as a custom attribute that this would break. Ch
114.
by vimax 4y ago
Do you need to know the individual state legislatures for every state as well? You’d only need to know those representing you, which would not change. The purpose of the senate is to have a fixed knowable number of state representatives, wh
115.
by vimax 4y ago
A recent approach along these lines is to do acoustic modeling of the vocal tract that would have produced the speech. Unsurprisingly, AI voices do not correspond to realistic human vocal tracts, for now. https://www.usenix.org&
116.
by vimax 4y ago
Always worth keeping in mind that Stanford protected John Yoo after authoring the torture memos, and it was MIT that pushed for prosecuting Aaron Schwartz. University administrators are as much your friend as HR. Edit: Berkeley not Stanford
117.
by vimax 4y ago
Imagine the glorious day when your new jacket comes with a monthly subscription!
118.
by vimax 4y ago
Check out Robert Sapolsky’s lectures https://youtube.com/playlist?list=PL150326949691B199
119.
by vimax 4y ago
On the flip side, why would you expect us to be equipped to observe more than 20% of matter give the narrow range of our vision and hearing? Anything that doesn't interact strongly at our scale wouldn't be useful evolutionarily.
120.
by vimax 4y ago
Nothing shows this like walking in Austin and seeing homeless musicians on every corner just as good as the live band you paid to go see.

More ›