dougabug

10 ms·

61.
by dougabug 4y ago
I don’t mean frontier as in “the American Western Frontier.” I meant frontier as in, “the frontiers of human knowledge.” It’s not clear to me that the latter is limited in the way that the former is. Creation of knowledge products both faci
62.
by dougabug 4y ago
This saturation of demand for software talent has been predicted for literally decades. In 1992, Ed Yourdon published “Decline and Fall of the American Programmer,” only to reverse himself four years later with the publication of “Rise and
63.
by dougabug 4y ago
Well, Levy won the bet, but just barely. More importantly, his dismissive attitude towards AI mastering chess (“[…] the idea of an electronic world champion belongs only in the pages of a science fiction book.”) was deeply shaken.
64.
by dougabug 4y ago
A wise old inventor told me, early in my career, that “people always significantly overestimate the rate of progress in the short run, and greatly underestimate it in the long run.” Text conditional image generation is still a fairly new th
65.
by dougabug 4y ago
The use of rolling shutter to increase effective sampling rate was present in the original SIGGRAPH 2014 Davis et al paper from Bill Freeman’s group at MIT, “The Visual Microphone: Passive Recovery of Sound from Video” ( https://d
66.
by dougabug 4y ago
Good point. Persistent brain fog shouldn’t be taken lightly, it might result from underlying (treatable) health issues. Regardless of what career I chose, the ability to focus and think clearly are fundamental to my personal quality of life
67.
by dougabug 4y ago
Learning math is a personal journey, it can be slow and difficult at times. No matter how smart one is, there will always be mathematics that one finds completely baffling. Generally, curiosity, motivation, patience, perseverance, maturity,
68.
by dougabug 4y ago
Scaling up improved versions of existing recipes can be done surprisingly fast if you have strong DL infrastructure. Also, GPT-3 was built on top of previous advances such as Google’s BERT. I’m surprised that it took Google so long to answe
69.
by dougabug 4y ago
That’s pretty speculative and dubious (the holding back part) given the heavy bias to publication culture at Google Research and DeepMind. OpenAI has hardly been “crushed” here; PaLM and Imagen are solid, incremental advances, but given wha
70.
by dougabug 4y ago
This characterization is not really accurate. OpenAI has had almost a 2 year lead with GPT-3 dominating the discussion of LLMs (large language models). Google didn’t release its paper on the powerful PaLM-540b model until recently. Similarl
71.
by dougabug 4y ago
Someone still wrote the prompts, perhaps iteratively until they got results which pleased them enough to save and share with others. Other tools allow modifying the results by coarsely sketching where you want semantic changes. You could ha
72.
by dougabug 4y ago
It depends on the level of quality, polish, precision or originality that you want. It’s pretty likely that some people will be much more talented and skillful with this new class of AI-art tools than others. People will figure out how to d
73.
by dougabug 4y ago
Sure, but they have tapped into a pre-existing fan base and many decades of storylines, themes, and characters. Disney has also done this successfully with the Star Wars franchise, and the Mouse House will likely continue to make most of it
74.
Emergent bartering behaviour in multi-agent RL(deepmind.com)
4 points by dougabug 4y ago|0 comments
75.
by dougabug 4y ago
Sure, but GPT-3 was trained by self-supervised learning on only static text. We see how powerful even just adding captions to text can be with the example of DALLE-2. GATO takes this further by letting the large scale Transformer learn in b
76.
by dougabug 4y ago
In machine learning, there’s a long term trend towards automating work that used to be done manually. For instance, ML engineers used to spend a lot of time engineering “features” which captured salient aspects of the input data. Nowadays,
77.
by dougabug 4y ago
Your position seems logically and morally indefensible. You accept that you individually and your society as a whole need both a (presumably effective) military and a defense industry. The latter must necessarily engage in significant defen
78.
by dougabug 4y ago
You agree that there should be a military and a defense industry, but anyone working on weapons R&D is a malevolent actor (ie “making the world a much worse place)? This seems illogical. The effectiveness of a military is tightly bound
79.
by dougabug 4y ago
> One person can control, at most, about two guns and one truck at a time. A really big gun, or missile (battery)…
80.
by dougabug 4y ago
But you presumably rely on your own nation’s military to help protect and safeguard your rights, freedoms, even your life. Without drone and anti-tank / aircraft missiles, Ukraine would already be lost, and Putin would be on to his nex
81.
by dougabug 4y ago
Prior to the invention of reliable, accurate firearms, for much of history a relatively small number of dedicated warriors (e.g. armored knights on warhorses, samurai, cavalry, archers, etc) could dominate a much larger population of common
82.
by dougabug 4y ago
Loyalty has go both ways. Let’s not forget that the tech industry not so recently conspired to hold down compensation by engaging in secret but pervasive “no poaching agreements.” Let’s not also forget that the tech industry has harnessed v
83.
by dougabug 4y ago
Machine learning, for sure. Pytorch. Andrew Ng has a new version of his machine learning class on Coursera, now in Python (the de facto lingua franca of ML these days). A game engine like Unity, or maybe Unreal Engine. People use game engin
84.
by dougabug 4y ago
All analogies are flawed to some degree. The objective is clearly simpler in the case of Go and other strategy game. Although recreating physical objects (either animate or inanimate) is in principle straightforward: you just need a certain
85.
by dougabug 4y ago
At some point, it will have to evolve beyond learning from human labeled (or even just collected) examples to learn by its own open ended dynamics and self-directed experience. This is analogous to the original AlphaGo bootstrapping from re
86.
by dougabug 4y ago
Uh, you’re not actually required to watch something just because it exists. Some subset of people not liking content isn’t, in and of itself, reason to stifle peoples ability to create said content. The job of a literature professor is not
87.
by dougabug 4y ago
To create specific people and spatiotemporal consistent content, we would use conditional image generation where we condition on both the text and some seed vector, and proceed autoregressively. If we wanted a random person, we would use a
88.
by dougabug 4y ago
Sure, but how is this different from all the blogs, vlogs, social media, and the Internet in general (with an assist from smartphones, digital cameras, game engines, camera drones, and real-time ray tracing GPUs) allowing humanity to create
89.
by dougabug 4y ago
Right. In the literature this is typically referred to as autoregressive sequence generation, and you can use it generate a whole story consistent with a given prompt (of any length), or even a whole image where you give it the first N seed
90.
by dougabug 4y ago
You can decompose an entire scene into a collection of 3D GANs, each of which is fit by GAN Inversion to the object instance which best explains what is seen in the images. Also, we absolutely can tell the difference between trash cans with

More ›