yayr

link to github model instructions: https://github.com/deepmind/physical_concepts

what usually helps is to include some of your assumptions with the effort estimate - especially the ones that cause significant effort.

In a product those typically are not so much negotiable with respect to the architecture. In a project they may lead the customer to change their architecture priorities completely.

with that reasoning they would need to allow Google or MS AD accounts.

Thanks for sharing this.

Regarding funding - I think this is a very crowded space and it is hard to see, how data station will get 10x better than other toolings. E.g. on the self-service side there are solutions like Alteryx, that are progressing quite quickly. On the dev oriented side e.g. for automated productive integrations you have various toolings available e.g. in the could platforms. Yes, they all cost money, but they deliver also value.

There will be a niche certainly for this tool somehow, but I think this path needs to be approached together with customers. So if you find someone who is willing to pay $$$ for a certain addition to the tool, you'll have some light into the direction this could head to be self-sustaining.

Be careful regarding taxes and "Betriebsstätten". If you are living in Germany and your company management is mainly operating in Germany you are liable still for local taxes ("Hinzurechnungsbesteuerung"). There are also special rules depending on how much of the company stock you own. Also you may have to register a "Gewerbe" in certain cases if you are doing business in Germany (e.g. selling something) from a foreign entity.

So in a bad case as majority owner you will then have to do the taxes and other admin topics in the country of your company AND in Germany and you'll always have a special relationship with a curious tax office.

So the only way to have really a startup in a non-German jurisdiction is to make sure that your management decisions (and usually also some infrastructure) is set up there.

Is it still worth designing new low/medium performance laptops that are not based on a SOC? I mean, the energy requirements for dedicated CPU/GPU units are just not good I guess

in case this can be deployed single-handed it might be useful on a spaceship... would need some relativistic time accounting though.

> Understand what the priority is, evaluate the ROI for the ideas you propose, determine whether the timing for it is the right one.

The issue with this approach is that most "lesser skilled" managers will not understand the ROI of fundamental work, that not immediately leads to an ROI, as sometimes those things cannot be well quantified. Those are the cases, where you then need to be able to trust the more skilled team members to make those decisions also on other value driving criteria.

I think, whatever you should do strongly depends on yourself and your circumstances. Burnout is mainly a stress response, but the way to find your balance again will really have to be your own way. It is great, that you discuss this already with some proper guidance. Whether your team and job can help you here, also depends. If that should be the case, it needs to happen in a way that reduces your stress. If you think so, why not try, but don't expect outsiders to give that final recommendation.

On that Netflix topic... You'd be much certainly better of simply trying to sleep or do relaxation exercises. Also trying to code during that time does not help relaxation. Try to establish clear boundaries between these different modes in your life.

If this stack gives the control over data back to the users, there will be some value. But I assume, even with this stack there will again be tendencies towards centralization driven by convenience, efficiency and simply bargaining power...

there is a reason, why tools like Tableau and Power BI exist. Excel fails at building stable solutions for presenting or modeling data. Actually, it also fails at stably and effectively preparing and transforming data for more complex scenarios.

It is great for freestyle. But people want to use it for more...

Can someone explain to me, what actually constitutes that a particle has information or has no information? I.e., what would actually cause a change of this state?

I would expect, that having just 0 or 1 alone would not. Maybe it is related to an observer/environment, i.e. the reduction of the wave superposition to certain possible states...

interestingly, most split air conditioning units also have a heating function. Usually, they convert 1kwh of electric energy into 4-5 times the heat energy.

Heating with electricity is therefore something quite achievable and sometimes even economical. At current German electricity prices of 0,33 EUR/kwh electricity heating is currently cheaper than gas heating (currently at 0,20 EUR/kwh for new contracts - was at 0,04 before the crisis).

thanks, C# 8.0 indeed seems to bring that, awesome.

https://docs.microsoft.com/en-us/dotnet/csharp/language-refe...

That's the key point. Since earth will not start charging directly for any environmental costs occured by its exploitation we need other rules for dealing with this. Market forces alone do not account for the price generations after us will have to pay for it.

One of the major innovations I miss still in C# is more flexibility regarding interfaces and inheritance. Like e.g. protocols in Swift. Their synthesized implementations are a great and productive method to ensure standardization and code reuse without much boilerplate.

C# is still on the old OO path coming from Java with respect to that and decorating something with multiple base implementations is still a pain.

most tools are multi-use. A smart phone is currently the best "convenient world remote control" you can have. It is also the best "distraction toy" you can have...

...self-discipline does not come easy, especially if supply and demand are strongly in favor of distraction...

If the assets seized so far by the international community (incl. most of Russias central bank assets) would be used for Ukraine reparations once this is over, this would amount to an immediate sum of more than 600bn USD as far as I know.

I wonder if Putin really is ready to leave all his other international assets behind for this.

it's actually 40 TEUR

https://www.dailysabah.com/world/europe/kyiv-offers-over-40k...

I like this blog to understand NVidia perf stats (search for "Normalized GPU deep learning performance"):

https://timdettmers.com/2020/09/07/which-gpu-for-deep-learni...

It seems that their consumer GPUs also have a gap

yeah, you could also say the performance is only 57% in this particular test for the M1 vs A100. Still not bad for a laptop though with a first gen chip.

there is also some hint about performance/watt vs an NVIDIA A100 using BERT models:

> The A100 draws 131W peak during the training run, the M1MAX GPU draws a maximum of 15.4 W. The A100 runs an iteration at 7ms vs 104ms (with SHARK) and 196ms (with TF-Metal)

2D panels in VR are interesting for me, but not a killer feature. What would be a killer feature would be a VR designer for blender, unreal, unity etc.

Blender starts going there, but not yet for SVR: https://docs.blender.org/manual/en/2.83/getting_started/conf...

I think, this would be a way to really differentiate from closed platforms (which I also expect Apple RealityOS to become).

Unfortunately, the expectation that easy to use apps and tools for creating and consuming VR content will only be available only on closed platforms is quite reasonable.

> You can use a web browser with all your extensions

Since WebXR is not supported yet, this breaks the value proposition for me currently, as all VR content delivered via browser will not yet work. As I said, once this issue is solved (I think they are working on it), I'd reconsider.

For me the linux desktop use case is too weak to spend so much money on a headset. Probably, the picture would change, if WebXR would be supported. This would introduce at least one widely accepted UI framework with VR scope to the platform.

What can you really DO in VR here in the future?

- VR videos?

- VR gaming?

- VR art?

:-D something like zkIP (zero knowledge internet protocol)

This case is about IP address exposure, not cookies. This would still happen. So everyone showing youtube videos would be affected unless users also start agreeing to IP exposure… this could probably be avoided by extending the sites terms.

So what happens if you include e.g. a Wikipedia image or a youtube video? Is that a GDPR violation too? These scenarios also lead to making the users IP available to a third party.

If so, how do we avoid breaking the web while keeping privacy needs in balance?

partially agree, but I also think that it is in many cases worthwhile entertainment compared to many other entertainment options available.

but also I think that understanding paradigms and thought processes outside your usual comfort zone is education, albeit not the formal one. And this is what Masterclass does great imho.

or this:

https://www.hardkernel.com/shop/odroid-go-advance-black-edit...

and it does come with a more digestible price tag