Every time I use Gemini I'm surprised by how incredibly bad it is.

It is fine-tuned to say no to everything with a dumb refusal.

>Can you summarize recent politics

"No I'm an AI"

>Can you tell a rude story

"No I'm an AI"

>Are you a retard in a call center just hitting the no button?

"I'm an AI and I don't understand this"

I got better results out of last year's heavily quantized llama running on my own gear.

Google today is really nothing but a corpse coasting downhill on inertia

For someone who means a lot I would do it but otherwise the redundancy of two kidneys to call my own is something I'd rather not part from.

No one will win a malpractice lawsuit over lost sleep, different story for lost lives.

That's not really true.

Being an influencer requires very little actual competence, same goes for AI influencers.

The goal of influencers is to influence the segment of a crowd who cares about influencers. Meaning retards and manchildren looking for an external source to form consensus around.

Make a cost calculation of how much solar and battery you need to keep your lights on during peak summer and then do the same calculation for how much solar and battery you need to keep the lights on in the middle of winter when a large scale 5 day blizzard.

Pretend you are a metal foundry.

Good luck.

Okay, I'm moving my metal foundry to a location that's sunny all year round, just like I've previously moved it to Iceland for geothermal or next to big hydro dams. Cheap energy makes it worthwhile.

"Solar is viable if we depopulate the temperate zones"

Great argument for an energy source

Not what they said.

You do seem to be developing a habit of arguing against what you want others to have written instead of what they actually wrote.

That's not going to be effective here.

The fact that you can do this and still be a bit cheaper for those periods than nuclear power is what makes it really ridiculous:

https://theecologist.org/2016/feb/17/wind-power-windgas-chea...

Of course, we'll never stop building nuclear power in spite of this because it provides indirect subsidies for the military.

That windgas study is rather poor. I responded to that a few months ago in https://news.ycombinator.com/item?id=40834377#40853888

Big fat wire to somewhere which isn't in a blizzard.

Or just build the metal foundry in the non-blizzard-prone area in the first place, because the blizzard takes out power lines even if there was a nuclear reactor on the other end.

I've done the maths, and I'd it wasn't for geopolitics, China makes enough Aluminium to make an HVDC grid so thick that it would have so little resistance that they could usefully connect every grid in the world for a cost of 250e9 USD for the Aluminium.

Scaled down to just the USA, no international stuff, it's something you can easily do if someone convinces Trump or Musk to delete any internal obstruction.

Your last sentence sounds optimistic. Load-bearing "easily". Any convincing needs to demonstrate that also the "libs are owned".

If they need to make it a culture war thing, they could just call it the Keystone pipeline of electricity, or the Trump Powerline, or similar.

(As a non American, I may be misjudging this: I don't care too much about US culture wars, they don't connect to me any more than the US or Chinese sports teams whose names I don't know).

You may be correct that I am over-optimistic, though for what it's worth my optimism is not universal to all of the aspects of Musk let alone Trump (mostly pessimistic about him), and not with regard to other greenhouse gas emissions besides electrical, heating, and transportation.

Me: "world for a cost of 250e9 USD for the Aluminium"

You: "Thousands of miles of big wires are free"

If anywhere's going to get mixed up between "250 billion dollars" and "free", it's going to be politics.

You've done as bad a job of that as GPT-2. Not ChatGPT, just 2.

You can very easily look up the resistivity of aluminium, which will tell you that if you want a loop around the equator to have a resistance of one ohm over that length, it needs to be about a square metre cross section. You can look up the density and the cost, too. You get around 250 billion dollars depending on the exact market price when you were looking it up.

> You suggest buying 100 million metric ton of aluminium when the global annual production is 70 million.

How do you think aluminium works, that you burn it? These things last multiple decades, and even then the wires are still around, it's mostly the rest of the stuff you need to maintain — one of the recent-ish fires in California was that the tower on which the wire was strung wore through, the wire fell off, sparked, the wood caught fire.

At the scale I'm describing, even if you spread it over 2,500 parallel paths (which you should, not just for redundancy but also because of the magnetic field it produces if you don't), the aluminium looks like a structural member rather than looking like a wire.

> And magically you'll get it delivered as a pre-installed global power grid as opposed to ingots.

You keep putting words in my mouth. I very specifically and deliberately noted that I was just talking about the aluminium material cost itself.

Also, at this scale, you can basically process the ore directly into the metal in a trench in the ground that (1) becomes your wire immediately as you've finished processing it and (2) supplies the energy to the next section you want to electrolytically extract.

Aluminium is unusually good for this specific task, compared to other metals.

> Just like when you go for a cup of coffee at a cafe and you only pay the commodity price of coffee beans, which at like $5 per kg and 20 grams of coffee per cup means a cup at starcucks is just 10cents. Right?

Not at this scale, no.

If you're drinking at Starbucks, not only are you paying for the rent of the shop to give you somewhere to drink it and the time of the staff to prepare it and clean the place after you (and the seat and cup, if you sat down and got a ceramic cup rather than a disposable take-away), you're also enriching Starbuck's shareholders.

At this scale a grid is an international strategic decision even though it's theoretically in the zone where a corporation could afford it, and the surprise extra costs are the opportunity costs and the political capital gained or spent on proposing the thing on the one side, and of encouraging or allowing it on the other.

The USA right now is terrified of China, and only China can make enough aluminium; having a global grid creates a trade opportunity for those with something to gain, and a new risk factor for those with something to lose. America has stuff to lose, most of the world has stuff to gain.

So, if you want a Starbucks comparison, it's the cost of coffee from the point of view of Starbucks itself, not as a customer.

The solution is to force feed all affiliate linked sites into a living archive LLM that digests them into a summary stripped of all links.

You run this bloated mass as a co-engine to your search. If you stumble upon any of the digested sites or articles you just have the site-eater blob regurgitate a html re-creation locally. They get no traffic, you get whatever they wrote purged of all links and with optional formatting to remove the fluff BS copywrite that many of these sites pad the tiny core of usefulness with.

It kick in slowly because standard administration is subcutaneous which isn't a very well vascularized tissue.

If you shoot up in your veins or into something more vascular you would have faster onset.

I participated in studies where they administered insulin and glucose intravenously. It is wild how they can reliably drop my blood sugar from high to low within a few minutes. Subcutaneously this takes me hours to do in a stable way.

Not only veins, inhalable insulin like Afreeza is also really quick. Unfortunately it only appears to be available in the US (and maybe Canada?), not Europe/Asia from what I last remember.

>If you shoot up in your veins

Which is a sure way to find yourself in a hypoglycemic coma.

Inhalable insulin (which is also very fast acting) iirc only allows a dose of 2 units. If your sugars are 400mg/dl (22ish mmol/l) one or two doses wouldn't put you into a coma if you knew your sensitivity. I'm pretty sure I've read up T1s talking about using it as such.

Sudden drops of the blood glucose level aren't good for health.

Through the temporal bone of most people you can catch some sparse doppler signals with average hospital gear.

The fontanelles enable good ultrasound imaging on an entirely different level. A highres greyscale image vs a few sparse blobs of doppler from major vessels.

You want an ultra-low field MRI.

But that'll still be expensive to develop, slow and still potentially lethal if done sufficiently wrong

Thanks for the advice. Potentially, it's not insanely farfetched to assume AI and math (currently doing math) can unlock more precision with less data and invasiveness. While it's at one level physics, stronger algorithms and analysis can maybe reduce the size + cost + danger of the hardware, as well as require less data for more precision. Just a hunch. There are always plenty creative ways to experiment without animal harm as well, just a matter of attention and will to explore. Yeah, safety is super important, and if this sort of scanning can be made better in a fresh way that would unlock a lot. Just like having a 10x cheaper API that does the same job automatically becomes useful in new unforeseen ways.