Technology

Machine learning and AI may be deployed on such grand tasks as finding exoplanets and creating photorealistic people, but the same techniques also have some surprising applications in academia: DeepMind has created an AI system that helps scholars understand and recreate fragmentary ancient Greek texts on broken stone tablets.

These clay, stone or metal tablets, inscribed as much as 2,700 years ago, are invaluable primary sources for history, literature and anthropology. They&re covered in letters, naturally, but often the millennia have not been kind and there are not just cracks and chips but entire missing pieces that may comprise many symbols.

Such gaps, or lacunae, are sometimes easy to complete: If I wrote &the sp_der caught the fl_,& anyone can tell you that itactually &the spider caught the fly.& But what if it were missing many more letters, and in a dead language, to boot? Not so easy to fill in the gaps.

Doing so is a science (and art) called epigraphy, and it involves both intuitive understanding of these texts and others to add context; one can make an educated guess at what was once written based on what has survived elsewhere. But itpainstaking and difficult work — which is why we give it to grad students, the poor things.

Coming to their rescue is a new system created by DeepMind researchers that they call Pythia, after the oracle at Delphi who translated the divine word of Apollo for the benefit of mortals.

The team first created a &nontrivial& pipeline to convert the worldlargest digital collection of ancient Greek inscriptions into text that a machine learning system could understand. From there it was just a matter of creating an algorithm that accurately guesses sequences of letters — just like you did for the spider and the fly.

PhD students and Pythia were both given ground-truth texts with artificially excised portions. The students got the text right about 57% of the time — which isn&t bad, as restoration of texts is a long and iterative process. Pythia got it right… well, 30% of the time.

But! The correct answer was in its top 20 answers 73% of the time. Admittedly that might not sound so impressive, but you try it and see if you can get it in 20.

The truth is the system isn&t good enough to do this work on its own, but it doesn&t need to. Itbased on the efforts of humans (how else could it be trained on whatin those gaps?) and it will augment them, not replace them.

Pythiasuggestions may not be perfectly right on the first try very often, but it could easily help someone struggling with a tricky lacuna by giving them some options to work from. Taking a bit of the cognitive load off these folks may lead to increases in speed and accuracy in taking on remaining unrestored texts.

The paper describing Pythia is available to read here, and some of the software they developed to create it is in this GitHub repository.

WeWork co-founder Adam Neumann didn&t plan for his familycontrol of WeWork to end at his death but instead expected to pass that control to future generations of Neumanns, too, says Business Insider.

The outlet reports that in a speech Neumann gave to employees in January of this year, footage of which it says it has viewed, Neumann is seen saying that WeWork isn&t &just controlled — we&re generationally controlled.& He reportedly goes on to say that while the five children he shares with wife Rebekah Neumann &don&t have to run the company,& they &do have to stay the moral compass of the company.&

According to BI, Neumann even invoked his future grandchildren, telling those gathered: &Itimportant that one day, maybe in 100 years, maybe in 300 years, a great-great-granddaughter of mine will walk into that room and say, ‘Hey, you don&t know me; I actually control the place. The way you&re acting is not how we built it,'& he said.

These may sound like more outlandish proclamations from Neumann, who has a flair for the dramatic. (Talking to Fast Company earlier this year, he compared WeWork to a rare jewel, asking, &Do you know how long it takes a diamond to be created?&)

But before WeWork began coming apart at the seams, Neumann had every reason to believe that he could pass power down to his heirs. Though many public shareholders may not realize as much, a growing number of tech founders enjoy the kind of dual-class shares that Neumann had extracted from investors, shares that don&t merely give founders more voting power for a while after their companies go public or even throughout their lifetimes, but whose power can be passed down to their children, too.

We wrote about this very issue as a kind of hypothetical last month, quoting SEC Commissioner Robert Jackson, a longtime legal scholar and law professor, who told an audience last year that nearly half of companies that went public with dual-class shares between 2004 and 2018 gave corporate insiders &outsized voting rights in perpetuity.&

Warned Jackson, &Those companies are asking shareholders to trust managementbusiness judgment — not just for five years, or 10 years, or even 50 years. Forever.& Such perpetual dual-class ownership &asks them to trust that founderkids. And their kids& kids. And their grandkidkids . . . It raises the prospect that control over our public companies, and ultimately of Main Streetretirement savings, will be forever held by a small, elite group of corporate insiders — who will pass that power down to their heirs.&

You might argue that itsenseless to worry, that the market will speak as it did in WeWorkcase. But not every company has such apparent flaws, and Neumann could have made himself a lot harder to shake than he did. In fact, the broader question the video raises is whether anyone will step in to stop the broader trend, or if public market investors will be living with the consequences down the road instead.

Neumann wasn&t insane to imagine the scenario that he did. That doesn&t mean itrational. Giving founders super-voting shares for some period after transitioning onto the public market, we can understand. Giving founders so much power that their kids call the shots of these publicly traded companies? Now that is crazy.

Despite early-stage virtual reality market and augmented reality market valuations softening in a transitional period, total global AR/VR startup valuations are now at $45 billion globally — include non-pure play AR/VR startups discussed below, and that amount exceeds $67 billion. More than $8 billion has been returned to investors through M-A already, with the remaining augmented and virtual reality startups carrying more than $36 billion valuations on paper. Only time will tell how much of this value gets realized for investors.

(Note: this analysis is of AR/VR startup valuations only, excluding internal investment by large corporates like Facebook . Again, this analysis is of valuation, not revenue.)

Selected AR/VR companies that have raised funding or generated significant revenue, plus selected corporates as of September 2019.

There is significant value concentration, with just 18 AR/VR pure plays accounting for half of the $45 billion global figure. Some of the large valuations are for Magic Leap (well over $6 billion), Niantic (nearly $4 billion), Oculus ($3 billion from exit to Facebook), Beijing Moviebook Technology ($1 billion+) and Lightricks ($1 billion). While there are unicorns, the market hasn&t seen an AR/VR decacorn yet.

Across all industries — not just AR/VR — around 60% of VC-backed startups fail, not 90% as often quoted. That doesn&t mean this many startups crash and burn, but that 60% of startups deliver less than 1x return on investment (ROI) to investors (i.e. investors get less back than they put in). To better understand whathappening in AR/VR, letanalyze the thousands of startup valuations in Digi-CapitalAR/VR Analytics Platform to see where the smart money is by sector, stage and country.