ryandamm

For what it's worth, film also suffers from reciprocity failure (nonlinear response at extended exposure times with low flux).

Some things were never easy.

This article is completely, utterly wrong as soon as it starts talking about anything plenoptic. Please disregard after that point as there are serious factual errors, particularly regarding what Google did or didn't do with Lytro.

(Source: I'm in the VR / plenoptic space, knew a bunch of people at Lytro, some of whom are now at Google. Timelines and facts do not match this article's assertions.)

Sadly not; the lens in the eye is somewhat adaptive, but the 'blooming' effect of bright, short-wavelength lights at night mean your eye is incapable of focusing all of the rays onto a single small spot.

Unless, that is, the bright light (color irrelevant) causes your pupil to contract. Then things will appear sharper and darker (until your pupil expands moments later).

It's sort of amazing that biological eyes work at all, to be honest.

This is due to dispersion, or ACA (axial chromatic aberration): your eye actually focuses at different depths for different colors. This will be most noticeable for large apertures (so at night, when your pupils are dilated), and for extremely long wavelengths... like highly spectrally pure indigos, like neon lights, or certain LEDs.

You're comparing VR to being in the car. The correct comparison is VR to a 2D screen, which is what this is replacing/augmenting.

It doesn't need to be the Holodeck to be useful.

The problem is focal conjugates; you can't darken something close to the eye and have it look like a dark region at a distance without focus blurring ruining the effect.

Pettis writes extensively about China and its internal debt problem. The best primer, though lengthy, is to slog through his posts dating back to the mid-2000s.

Briefly, though: China has had artificially cheap credit in its corporate sector due to explicit government policies restricting the growth of household spending. This cheap credit caused a lot of malinvestment ("bridges to nowhere"), and there hasn't been an accounting reckoning yet.

To continue to grow, China needs its household sector to grow, both in consumption and share of the economy. This is a classic shift from middle-income to developed economy. Because right now continued growth from the corporate (generally export-driven) sector has topped out, and getting more growth out of it is costly in terms of debt.

I think those are the broad strokes. To summarize: the Beijing is responding to the explosive growth of internal debt in the economy, and the goal is to continue growing. That growth must come from the household sector, but must also include dealing with the massive debt overhang. (I suppose you could also say that Beijing's goal is to stay in power, which requires economic growth to keep the populace happy.)

With all due respect, interpolating light fields is far less trivial than you make it out to be. It's a 4D field, and naive interpolation leads to loss of detail, and often edge doubling (itself a form of aliasing).

Furthermore, if you're interpolating rays, you're necessarily not doing what you originally proposed, which is to only light up a (random or pseudorandom or evenly distributed) subset of the pixel display elements, presumably to save on rendering cycles.

Let me just say, more generally, that intuition trained on 2D doesn't apply directly to light fields.

If you can drive the underlying display, you can drive the Looking Glass. The pixels you're lighting up just get distributed across views.

The videos are pretty accurate to the experience, having seen them in person a couple times. I don't think there's any cinematic trickery going on.

That's an awfully crude test, that derives its power from its appeal to emotion.

Which is not to say it's prima facie wrong, but to point out that its correctness should be separated from its emotional impact, which is very tough for most humans to do.

That assumes your pixel distribution has much higher potential addressable resolution than the signal you care about... in which case, why not just display the full raster?

Rendering for 2D and display are two different beasts -- but I'll own the fact that I'm not formally trained in this and there may be subtleties -- or even obvious signal processing facts -- that I'm getting wrong. (But if I'm wrong, I'd love to know how, for my own edification.)

If you don't display the full raster, you're aliasing the signal and will see a lot of view-dependent artifacts (twinkling).

Source: I work in light fields.

I skimmed the repo. Setting aside its (probably repugnant, charitably just juvenile) politics, it is objectively not funny.

Like, seriously. I'm annoyed I spent 30s scrolling through it. And it reinforces my belief that conservative thinking and humor are completely incompatible.

I used to dismiss those obvious Hollywood cliches, but the more I looked into the space professionally, the less silly it sounds. Images-as-data are far richer than images-as-images, it turns out, and clever algorithms can extract a lot more information from them than you would naively expect.

So yeah, you're right!

This is traditionally called a histogram in video. It allows you to see brightness by region, though it's typically stacked vertically.

Visby | Software Engineering | San Francisco, CA | Full-time | Onsite

Visby is building a new imaging standard for natively 'holographic' displays, like VR, AR, and multi-perspective screens.

Our tech is based on the concept of 'light fields,' the set of all light rays passing through a space. By capturing or representing all the light rays, you can determine what a user would have seen from any possible perspective. This is superior to traditional, 3D model-based imaging, because it doesn't look like a video game. Our approach is more photographic than other approaches, and leads to realistic images. While most of the work is software, we do have a huge pile of cameras and work closely with the film industry.

The problem space is very complex, involving the latest math and algorithms, and combining ideas from a huge, dizzying number of technical fields. So you don't need to be an expert in everything to apply, just have a passion for hard problems, be willing to work on a team, and be excited to learn (and teach). We have three open positions, but we will try to build a role for the right candidate.

Algorithms Engineer, Light Field Imaging: https://jobs.lever.co/visby/5f861491-fe0f-4003-99ab-01f9456a...

Research Engineer, Light Field Imaging: https://jobs.lever.co/visby/5ae6b083-ffec-4b1e-89db-8d81342b...

Software Engineer, Virtual Reality: https://jobs.lever.co/visby/662abcfa-6d30-4844-951a-75f1796e...

Personally, I doubt it barring exceptional circumstance: previous achievement in another field, metrics that can't be ignored (tough to get via bootstrapping, though)...

I have heard from some seed stage investors that they will do priced seed rounds -- even some that claim to prefer it -- so there may be technical exceptions that really reinforce your point.

To be fair, the workshop's website also asks about previous fundraising. I've found this always comes up during fundraising discussions, too. Partially because the company's finances matter, but also because investors want social proof. I don't think this is as negative as it seems: early stage startups are black boxes, and any signal is helpful, I suppose -- and there's some sense that previous ability to 'sell' is indicative of future success. (I try to empathize with the investors on this one; they see a lot of pitches.)

Hi Justin,

Thanks for the article, and putting on a workshop on fundraising. Regarding the workshop -- it seems like the application's only substantial question is "what business metrics do you have to support your beliefs?"

Do you believe there are quality businesses out there who can, should, and will be able to raise an A that don't have obvious metrics? It seems like that sort of question only applies to website- or app-centric companies that have obvious and easily-anagrammable metrics (ARPUs, CAC, churn, etc).

It's my feeling that some of the best, most disruptive companies won't have those metrics if they're operating in another space. (Though I have my bias; my company is one of those, at the moment.)

I fear that spreadsheet-driven investing tends to overfund incremental companies, but won't make the foundational investments required to build new platforms. But again, that's my bias.

Curious for your thoughts -- thanks!

Alibaba.

Most other shops will be expensive, occasionally you can find deals on anchor optics clearance (though that info is years old, and to have to know what you're buying!).

You're almost always better off buying the cameras and sensors together, unless you have specific needs. Sadly, this isn't something where you can save money by doing it diy. Supply chains are good and overall demand is too high for deals outside the normal channels, I think.

Would love to be proven wrong, but I used to pay around with optics regularly and it was always more expensive than I liked. Cutting open photographic lenses was a common way to get fun glass for cheap... So getting cheaper glass is tough.

I don't think your analysis is sophisticated enough to capture the actual underlying dynamics.

Rating of debt is largely a question of future ability to repay. The concerns of the rating agency, I imagine, is that much of the debt in China is being created for non-economic investment, that ultimately won't be able to pay for itself. And someone will have to foot the bill, whether it's the lender (via a write-down or write-off of nonperforming loans) or someone else (the central government, by assuming the loans directly or some other mechanism).

Yes, the Chinese central government has a lot of capital to play with, due to a very high household savings rate and positive trade imbalance. But the magnitude of the internal debt, and the degree to which it's likely misallocated and ultimately nonperforming, should concern anyone who's paying attention.

This explains why there aren't external forces that can check the growth in credit.

But there has been an explosion in debt within China, financing ultimately non-productive investment. Even with a positive external debt outlook, you can get wild economic imbalances within a country via uncontrolled credit creation. The hangover could be profound.

To put it another way: why do you think centrally-planned economies are frequently outpaced by capitalist economies? Misallocation of resources. Debt binges are similarly distorting.

Excellent, as an American I wish them all the luck in setting up an alternative currency standard.

The US has been hurt by the 'exorbitant privilege' of having its currency be an international reserve standard. Dollars that pile up on balance sheets and central banks' books get recycled into US assets and debt. This is part of the dynamic that drives the trade imbalance, which (on net) imports unemployment into the US, via lack of competitiveness.

Yes, the dollar would drop in value if the world's banks used a basket of currencies instead of large, dollar-denominated holdings (often T-bills). And that would be good for the US.

For a lot of these transactions, the mortgage isn't for the full price of the home. Partially because buyers have other assets (like stock, or cash that needs to be parked somewhere).

But also because the mortgage interest deduction tops out just shy of $1m, I believe. So the government-sponsored giveaway (being able to use pre-tax income to pay for your mortgage interest) tails off quickly, and the rational calculus around loan-to-value shifts dramatically.

For all the comments about listing shenanigans, we're overlooking a basic game theory issue here: when you set a price that looks to be near the value of the home, you're saying you're willing to accept reasonable offers. You're setting an implicit cap on the value of the home.

An obviously undervalued home, though, in addition to driving up interest (and creating a bidding war), suggests a much fuzzier cap so there's less signaling around price and a less obvious ceiling to offers. In other words, buyers are stuck wondering what other buyers might offer, rather than discussing whether to go some normative amount over the ask (like 10-20%).

Also, there's a 'curse of the winner' dynamic in auction-like environments, where the deal will necessarily go to the party that values the asset the highest. (This is evidently destructive to profits in oil and gas leasing auctions, for what it's worth.)

I think that's the dynamic, because no one is actually fooling anyone by underlisting prices, and real estate agents talk to each other and know immediately if something is priced to generate interest, or if the price is closer to what the buyer expects.

I assume that's last "spring," and not that your team is doing agile real estate sprints. But I'd be happy, and rather curious, to be wrong....

Not on a 13k square foot lot!

But yeah, that has happened elsewhere in the Bay Area. (Zoning tends to prevent it, though, so there's an element of regulatory arbitrage.)

I doubt that's the "usual" case -- yes, cash-only transactions occur, and foreign investment buyers are in the market, but I don't see a ton of vacant homes. I recall hearing from a real estate agent that foreign purchases were fewer than 10% of the transactions, but I have no idea what her source was.

Vancouver circa 2016, though...

Let me just say "thanks" for your help with our immigration needs!

Because Peter is probably too modest to self-promote, let me do it for him: working with him is great, it was completely friction-free and we got our employee's visa situation handled very, very quickly. Highly recommend.

Don't forget Daqri: https://daqri.com/

Enterprise is the one vertical that's still pretty strong in VR/AR, perhaps because sky-high unit costs and dev investment can be amortized more easily.