jmmcd

31 ms·

241.
by jmmcd 7y ago
Please use an informative title.
242.
by jmmcd 7y ago
Interesting to see McEwan mentioned as present (though he doesn't speak in this transcript) -- a great writer, with a nice recent novel about AI, but not really a technical guy!
243.
by jmmcd 7y ago
> One part of your talk is saying there is this range of metaphorical domains—dynamic systems, control systems, biological adaptation, resonance models—different kinds of pictures, and of that panopoly, we’ve chosen the computational alm
244.
by jmmcd 7y ago
Please choose an informative title.
245.
by jmmcd 7y ago
They are beautiful regions. Have you had the chance to visit?
246.
by jmmcd 7y ago
I was thinking more of UK (the only place where "pip pip" makes sense as a stereotype) versus the rest! Are you saying there are those in the UK who say "negative five"? That would surprise me.
247.
by jmmcd 7y ago
Anyone who says 'zed' and indeed cheerio, pip pip, won't say "negative five". They'll say "minus five", of course.
248.
by jmmcd 7y ago
Theorists of creativity eg Margaret Biden don't limit it to aesthetic domains. They include creativity in Go, in mathematics, even in coding. Marcus Du Sautoy has a recent intro book.
249.
by jmmcd 7y ago
This is only for inference, which is very cheap already, cheap enough for most applications (real-time video processing an exception). Training is the slow part which is worth putting on a GPU.
250.
by jmmcd 8y ago
Great article! This is total nonsense though: "Virtually all contemporary AI artists credit Mordvintsev’s DeepDream as a primary source of inspiration for their interest in machine learning and art."
251.
by jmmcd 8y ago
Yes, and the vast, vast majority of important papers by the three winners were co-authored in the same way with supervisors, graduate students or other collaborators.
252.
by jmmcd 8y ago
I agree that a "voting" mechanism sounds all wrong, not only for the reason you said. In the scenario described, where we have some neurons coming ultimately from the hand and sense of touch, and others coming ultimately from the
253.
by jmmcd 8y ago
What a set-up!
254.
by jmmcd 8y ago
Even though the wording is a bit ambiguous, the problem is to output an array of neighbour-sums, not a single sum-of-neighbour-sums.
255.
by jmmcd 8y ago
To be more explicit, I see "how does it know?" as analogous to code like this: if (condition == True) return True else return False
256.
by jmmcd 8y ago
I applaud this effort, but it's still not hitting the right notes, eg: > how to show the changes of a file in the staging area: git diff --staged Ok, but this shows the diff against what: the file in the working directory, the file
257.
by jmmcd 8y ago
> Firstly - cancel article 50, then reenact it again. That is within EU rules and avoids needing any yay or nay from EU. This is not possible. The recent judgement that UK could unilaterally cancel A50 was predicted on a "good faith
258.
by jmmcd 8y ago
A lot of people are saying hang on, they're useful for explaining. But my experience when teaching relational databases is that they give the illusion of understanding. I'd rather have the students know that they do not know (an
259.
by jmmcd 8y ago
> it's the priority of 99.9% of those doing slide presentations No, it's definitely not. Most people I see (I see a lot) are presenting from their own laptops. For the record I'm not involved in this project.
260.
by jmmcd 8y ago
This is a classic HN comment. Contempt which turns out to be grounded in "I have an esoteric priority which is different from that of the author." Anyway, it exports pdf.
261.
by jmmcd 8y ago
The extravagance brings the potlatch [ https://en.wikipedia.org/wiki/Potlatch ] and similar customs to mind.
262.
by jmmcd 8y ago
See also the tutorial slides available here https://www.cs.ucr.edu/~eamonn/MatrixProfile.html
263.
by jmmcd 8y ago
Yes, Google Scholar is so useful and ubiquitous that it is dangerous. It would be far preferable to have a proper open replacement for it. The technical issues are probably "easy". The ongoing maintenance costs/hassles could
264.
by jmmcd 8y ago
In the end, they're optimistic that Twitter can be a positive and fun experience, but I don't share that optimism.
265.
by jmmcd 8y ago
And ironically, Push/Clojush don't actually use trees! I quite like Push, and also Linear GP, the DAG-based system used by Schmidt and Lipson, and grammatical GP, depending on the application.
266.
by jmmcd 8y ago
I've a fair bit of experience with both GAs and LAHC and in my opinion the results are mixed. LAHC wins for fewer hyperparameters (1 versus a minimum of, say, 3), but in terms of implementation simplicity there isn't that big a di
267.
by jmmcd 8y ago
> Think of it as a logic search engine embedded in a general-purpose programming language, because that's what it is. I think this is the whole problem. The Prolog language is not great as a general-purpose language, and the "l
268.
by jmmcd 8y ago
But I totally agree! We shouldn't even write code for a decision table, because it's generic code which doesn't need to be adapted at all. Just use a decision table implementation off the shelf. But I don't agree that en
269.
by jmmcd 8y ago
> Good specifications are often written at a level that generating executable code from them is often undecidable and almost always intractable in practice [...] (which we don't know how to do just yet). Sure, but I'm talking a
270.
by jmmcd 8y ago
I agree, but what's even better than a specification is an executable specification. If we see a decision table as data (see discussion below) then it can be both -- no chance for the implementation to go out of sync with the spec, an

More ›