Yeah, this. We just put the genes to create whatever molecule we want in a bunch of e. Coli and grow them in vats then harvest the molecule.

It's super easy to put genes in e. Coli since they don't have a nucleus, you just shove the DNA in there and away they go.

Yes, blood is toxic to neurons. We have a blood-brain barrier that only lets some things through (either passively if they're small enough like oxygen or alcohol, or actively for bigger molecules).

The brain sits in cerebrospinal fluid (CSF) rather than blood.

As for the actual mechanisms that blood will damage the brain -- first, if you've got a brain bleed, it'll increase intracranial pressure and make it harder for new oxygenated blood to get in, which will kill your neurons from simply lack of oxygen. Second, even a small bleed will kill neurons, because things like the sodium concentration in your CSF will get messed up, and then when the blood dies in your brain it'll leave behind a bunch of iron from hemoglobin which will also kill your neurons.

I think it depends on the framing. Do we pick certain people to keep, or is it random? Do they all just disappear and get Raptured one day, or do we have some time to prep?

I'd guess that we could probably get by with a few million without a massive drop in quality of life, but that's if we had time to prepare. Wiping out 99.99% of the human population all at once would not result in a good time for the survivors.

OS. Kernel. Bus. Processor. Transistor. p-n junction

>Is the output from that antenna a variable waveform photon, or is it multiple photons of different wavelengths being produced at once.

Not just multiple, preposterous numbers of photons are output by radio transmitters; numbers like 10^34 photons per second

The variable waveform isn't from approximating the signal with a series of sine waves like Fourier analysis. It's from changing the wavelength of the photons you're spewing out (in the case of FM), or from increasing / decreasing the number of photons you're spewing out (in the case of AM)

Oh I meant fitting to the statistics of ImageNet / the training dataset. There's always got to be some kind of normalization

With data augmentation techniques (especially contrast or luminance randomization), normalizing would end up being a no-op in the end, right?

trump's name for him is "DeSanctimonious" since he's been using the same "I'm Jesus" religious iconography that Trump does

I mean it's the current reality.

AI meant something very, very different 30 years ago than it means now. Massive commercial success in a few areas will do that.

I think of AI as a subset of ML, focused on deep learning theory, architecture, and application.

We understand that, but the endocrine system which evolved before Cheet-os doesn't.

I know some companies that haven't hit their hiring targets for ML, but they're in the B2B space

That's just VC funding though, right? There's still a ton of investment from the big co's