111–115 of 115 posts
Earlier quoted context omitted.
Isn't the main point of the singularity that AI would be able to break free of such bonds due to its superior intelligence?
> Isn't the main point of the singularity that AI would be able to break free of such bonds due to its superior intelligence? At this point, the singularity is speculative fiction.
The idea of a virtual anchor was pioneered >10 years ago by a group at the Northwestern CS Department including Kris Hammond. The project was called "News at Seven." This video shows the later iterations of it: https://www.youtube.com/watch?v=M3MRkFYM9Q4 IIRC the early version around 2005 or 2006 used the Half Life 2 engine, so you would have Alyx Vance reading the news with a gun strapped to her belt. At the time I…
I can cite an earlier example, perhaps?
Back around '98 I was part of a 3d game project that had a dynamic news reader that moved his lips in accordance with a wav file being played. So, although 'not AI', it did do a realtime FFA on the waveform and 'guestimate' the shape of the mouth based on significant frequencies. Analysis of volume in the wav would trigger other movements (rocking backward), or increasing pitch might trigger eyebrow movements, etc. I recall that the 'wav' itself was stitched together from multiple wav snippets at runtime to give an accurate account of the game play that had just happened.
At the end of it all, it definitely satisfied the 'virtual anchor' as described here, and was all achieved with simple heuristics.
That said, I think I would've enjoyed being part of the virtual anchor team in the video.
Earlier quoted context omitted.
Once general-purpose GPUs get 100x faster (by 2025?), they'll get classified as weapons and one would need a strictly regulated license to do Deep Learning on them, with computations logged/snapshotted onto some governmental server. Gamers will be forced to stream from cloud.
Can you name another common computer technology that has gotten so advanced it became classified as a weapon? The notion that in six years general purpose GPUs will be strictly regulated as weapons, is pretty comical. It'll never happen. In three or four years the improvemnet scaling will already be enough that whatever scenario you think must wait six years to justify strictly regulating GPU ownership, will already…
Earlier quoted context omitted.
I'm not confident that's an issue. The press usually takes the party line anyway. In any case where the press feels compelled to not take the party line (it's happened now and then), they've spoken out. I think if the press felt it necessary, they could do a "special" presentation when they feel necessary. Honestly, I don't see a lot of difference between them and the way Sean Hannity blatantly takes the Trump line.…
It would be very hard for anyone to watch most of Fox News and not see it as some form of state-controlled media in this administration. I guess my brain immediately goes to the situation you describe with little/no human oversight. If the only layer of human oversight that does exist values control over truth, this is just one more layer of human intervention removed.
Wow text-to-speech doesn't seem to have gotten very far in the last 20 years? When we were kids we used to use a windows (3.1?) program for prank calls and it didn't sound much worse.. I would have thought by now basic things like word spacing and tones would have improved to the point where you would only realise its artificial occasionally.
It definitely doesn't use state of the art TTS. It sounds similar to MBrola english voice. There is google's [Wavenet]( https://deepmind.com/blog/wavenet-generative-model-raw-audio... ) and even chinese Baidus [Clarinet]( https://clarinet-demo.github.io/ ) which are much better.