February 24, 2019

Pritam Gupta - page 1480

Pritam Gupta has 7607 articles published.

Microsoft Brainwave aims to accelerate deep learning with FPGAs

in AI/Artificial Intelligence/chips/Cloud/computing/Delhi/Developer/Enterprise/fpga/Hardware/India/Intel/machine learning/Microsoft/Politics/TC/Technology by

This afternoon Microsoft announced Brainwave, an FPGA-based system for ultra-low latency deep learning in the cloud. Early benchmarking indicates that when using Intel Stratix 10 FPGAs, Brainwave can sustain 39.5 Teraflops on a large gated recurrent unit without any batching.

Microsoft has been pouring resources into FPGAs for a while now, deploying large clusters of the field-programmable gate arrays into its data centers. Algorithms are written into FPGAs, making them quite efficient and easily reprogrammable. This specialization makes them ideal for machine learning, specifically parallel computing.

Building on this work, Microsoft has synthesized DPU or DNN processing units into its FPGAs. The hope is that by focusing on deep neural nets, Microsoft can adapt its infrastructure faster to keep up with research and offer near real-time processing.

For as retro of a technology as FPGAs are, the nifty chips have been having something of a renaissance in recent years. Mipsology, an FPGA-obsessed startup has been working closely with Amazon to make the technology accessible over Amazon Web Services and other platforms.

If the last few decades were about creating generalized CPUs to handle a broad array of processes, the last few months have been about building custom chips that excel at specific tasks. A solid chunk of these efforts have gone towards dedicated chips for machine learning.

The most notable effort by far has been Google’s Tensor Processing Unit. The chips are optimized around TensorFlow and early benchmarking completed has been very promising. That said, most major tech companies have side projects around the future of computing — quantum chips, FPGAs etc. And when there’s interest from tech giants, startups are willing to make a game of it — Rigetti, Mythic and Wave to name a few.

No word on when Brainwave will be available for Microsoft Azure customers. As of now the system works with Google’s popular TensorFlow and Microsoft’s own CNTK. The company is setting ambitious expectations for deep learning performance using the technology so there will undoubtedly be plenty more benchmarking news in the coming months.

Featured Image: Billy H.C. Kwok/Bloomberg/Getty Images

News Source =

MIT’s Robogami lets you build custom 3D-printable robots from standard, folding parts

in CSAIL/Delhi/Gadgets/India/MIT/Politics/robotics/Science/TC by

Flat-pack furniture made Ikea a global powerhouse, and the same principles may help create a new generation of robots. Interactive Robogami is a project from MIT that lets users create ambulatory robots from a library of pieces that fold and fit together like origami.

“The goal is to make the process of designing robots accessible,” said Adriana Schulz, a PhD student at MIT who co-led the project. “The actuators, the materials, the code, things like that require a lot of knowledge. Our system encapsulates that expert knowledge, so the user can focus on conceptual design.”

Robogami lets users combine a library of intercompatible parts with primitives that can be printed like puzzle pieces and then folded and locked into shape. The flat-printed style reduces both print time and material cost by more than half.

It has a user-friendly GUI that has more in common with a game or 3D doodle app than a CAD or other design environment. You drag the pieces where you want them, arrange the type and placement of the feet or wheels, and add any other features you think might be useful (or cool-looking). Then you can work out how and when those parts will move, what direction, and so on.

Meanwhile, the app is doing all kinds of calculations in the background.

“One of the key things is that you can design geometry and motion at the same time. Normally that’s two different processes, but here you can change one and see how it affects the other,” said Schulz.

“The ordering of the how the legs move, the speed, these are all parametric structures,” she said. How they interact is systematized and you don’t have to worry about, for example, synchronizing motor torques or timing a foot’s touchdown — it’s being calculated internally. “The software maps these components into a full fabrication plan, from the mesh that goes to the 3D printer to the motors and code, the user doesn’t have to worry — it’s very end to end.”

For instance, if you decide to move the front legs back towards the center, the software calculates how that might affect when and how fast they move, or whether their motion would interfere with another piece. It watches for things like wobbliness, changes in orientation, and center of gravity.

  1. oriinterface

  2. variety

  3. ori_print

  4. ant

  5. IMG_5525

It might sound a little like baby’s first robot kit, but it’s clearly a versatile tool. In a handful of tests, Robogami users of varying levels of familiarity with CAD and engineering tools, “We were really surprised by the diversity of models people made,” Schulz told me. It took about 15-30 minutes to design a robot, a few hours to print it, then 30-90 minutes to assemble it.

The tool isn’t aimed at any population in particular, Schulz told me (other than “anybody”), but I thought this would make for an amazing couple weeks in a high school or college engineering course, perhaps even earlier. Design your own little bots, print them overnight, assemble them to understand how the motors and chassis fit together, then race them or have them navigate obstacles.

Right now Robogami is just the subject of the researchers’ paper, now published in the International Journal of Robotics Research. But Schulz said the goal was to move beyond mere locomotion and into other tasks. “Empowering people to design complex things is difficult! But I think it would be exciting to create tools that lower design barriers for casual users.”

“These tools enable new approaches to teaching computational thinking and creating,” said Daniela Rus, director of MIT’s CSAIL and collaborator on the research, in a news release. “Students can not only learn by coding and making their own robots, but by bringing to life conceptual ideas about what their robots can actually do.”

Featured Image: MIT CSAIL

News Source =

Blue Apron implements a partial hiring freeze and lays off 14 recruiters

in Blue Apron/Delhi/India/Politics/TC by

Blue Apron is chugging along on its post-IPO bumpy road. For starters, The meal-kit company’s VP of human resources and talent, Kate Muzzatti, has left the company, Bloomberg first reported. In light of Muzzatti’s departure, Blue Apron is looking for a chief human resources officer, which is an entirely new role at the company, TechCrunch has learned.

Blue Apron has also put into effect a hiring freeze on some salaried positions, but is still hiring for certain corporate roles and hourly fulfillment center positions. To wrap up the trifecta of bad news, Blue Apron let go 14 members of its recruiting team.

“As part of our long-term strategy, we recently completed an internal reorganization which involved several changes to our organizational structure, including the creation of our new consumer products team and the launch of a new fulfillment center team in Linden, which will replace our prior center in Jersey City,” Blue Apron CEO CEO Matt Salzberg said in a statement to TechCrunch. “As part of these changes, we temporarily paused hiring for certain positions, while keeping others open, for the duration of our 2018 resource allocation process. As a result, we implemented a small reduction in staff on our recruiting team. Like any company, we carefully assess ourselves on a regular basis and implement changes as needed to manage our business for the future.”

Since going public, Blue Apron has been hit with a couple of lawsuits — one from a former employee alleging violation of the Family and Medical Leave Act and a couple from law firms alleging Blue Apron misrepresented the challenges it faces with customer retention, delayed orders and reduced ad spend.

News Source =

This 23-year-old just closed her second fund — which is focused on aging — with $22 million

in Delhi/India/Politics/TC/Venture Capital by

Laura Deming is not your typical venture capitalist. Then again, she isn’t typical in many ways.

For starters, the 23-year-old, New Zealand native was home schooled, developing a love of math and physics and, perhaps most interestingly, the biology of aging, along the way. In fact, she became so preoccupied with the last that at age 11, Deming wrote to Cynthia Kenyon, a renowned molecular biologist who specializes in the genetics of aging, asking if she could visit Kenyon’s San Francisco lab during a family trip to the Bay Area. Kenyon said yes. When, soon after the visit, Deming asked if she could work in the lab, Kenyon said yes again.

Deming’s family moved to the U.S. to make it possible, and it’s highly doubtful they regret the decision. Indeed, by age 14, Deming was a student at MIT, and two years after that — at the tender age of 16 — she was a college drop-out, having been accepted into Peter Thiel’s two-year-old, Thiel Fellowship program, which gives $100,000 to young people “who want to build new things.”

Often, those “new things” evolve along the way. Not for Deming, who pitched the idea of a venture fund that would support aging-related startups, and has since turned that early concept into Longevity Fund, an early-stage venture outfit that just closed its second fund with $22 million.

Earlier today, we caught up with Deming, who’s now 23, to learn more about her path — and which technologies she’s betting on to extend the human lifespan.

TC: It’s incredible that this all started with an email to a UCSF professor.

LD: [Cynthia Kenyon] is the most amazing person I’ll ever meet.

TC: What did you do in her lab, exactly?

LD: We were working with tiny worms that are see-through. You put them on plate of jelly and you see what happens if you change their genetic material. Do they live longer or die faster? If you starve them, they live longer. If you starve worms and also turn off certain genes, could you get them to live even longer? I was naïve, but I really wanted to make the longest-living worms ever. [Laughs.]

TC: What did you study at MIT?

LD: I majored in physics actually, but I continued to work in a couple of labs, including [one overseen by] Lenny Guarente [a biologist known for his research on life span extension]. It was a lot of fun. I thought I’d be a scientist, but a grad student familiar with the Thiel fellowship told me I should apply and I did. It’s funny, one of the directors of the [Thiel] program told me recently that he thought I’d fail, even though he was very supportive. After we closed the first fund, he was like, “I never thought that would work out.”

TC: Why?

LD: In part because not long ago, if you talked with most VCs about aging, they didn’t think there was anything there. I think aging is such a young science, they hadn’t heard about it. Meanwhile, I care a lot about it, and though we don’t know if it’ll work or not, it’s not unlike [biotech companies trying to tackle] cancer in that way, and if you believe in cancer companies, you should also care about aging companies.

TC: How much did you raise for that first fund?

LD: A grand total of $4 million, and I was very proud of this. To be honest, I’d assumed $100,000 was enough to build a fund until I arrived in San Francisco and realized it was really enough to live on for two years. When I started fundraising, I was 17 — too young to legally sign contracts. I’d never managed money before. But I could talk to people about the science and got them on board with that. In the end, we had great anchor investors come together, and we invested in five companies that kind of proved out the strategy.

TC: Were one of those anchor investors Peter Thiel?

LD: We don’t really talk about our LPs.

TC: You say “we,” though you’re the sole general partner of Longevity. Is that correct?

LD: Yes, but I have a lot of back-office support. The way Longevity is structured, I’m also able to pull in the best people who have expertise from different domains, so it’s not one person who looks at all the deals.

TC: And these advisors get a stake in the company?

LD: Sometimes. Others — especially grad students — like to be paid up front. We’ll find the best incentive for that individual and work with that.

TC: One of your portfolio companies is Unity Biotechnology, a company that’s trying to reverse aging through therapeutics. Didn’t it just raise a giant Series B round this week?

LD: It did. All of the companies in that portfolio have [at least] raised Series A rounds of $30 million or more to get to that proof of concept.

TC: Given the amounts involved, is the plan to form special purpose vehicles, or SPVs, around your break-out winners?

LD: We like to help LPs follow on, so we look to do that in whatever way makes sense for both parties. With Unity, we put in money as early as possible because Ned Davis, who runs the company, is amazing and we thought its aging thesis would succeed.

TC: How many companies do you expect to fund with your newly closed fund?

LD: Eight to ten companies.

TC: Do you think your work will be harder, given that investors seem to be paying much more attention to aging suddenly?

LD: No. With our first fund, we spent up to six months with each deal, tracking the company before it was even raising. It’s something LPs really value from us; they know when they invest in something that they don’t need to re-do the diligence, that we’ve already looked at a bunch of stuff and we know this is the best possible investment in [a particular vertical].

Earlier, our biggest challenge was getting other investors on board and convincing them that aging has become a place to play. Now that’s a non-issue, which is great. Our job is to help the companies get other investors on board, so it’s wonderful to see excitement in the space begin to build.

TC: You look at a lot of technologies. I have to ask: do you find these new blood transfusion startups as interesting as the writers of HBO’s “Silicon Valley”?

LD: [Laughs.] While scientifically interesting, I think they get a little over-discussed in the press because of that vampiricism. It’s not as sexy to talk about new genetic regulatory elements that control the aging process. That’s not going to get as many clicks as a story about drinking the blood of your five-year-old.

News Source =

Salesforce slides past its $10B annual run rate target

in Apps/Delhi/earnings/Enterprise/Finance/India/Politics/Salesforce by

Earlier this year it looked as though Salesforce was starting to position itself as a company that would hit $10 billion in annual revenue run rate later this year — and, if all seems well, it looks like it may hit that target.

Salesforce reported its second-quarter earnings today, saying it brought in $2.56 billion in revenue on earnings of 33 cents per share. This is a beat on both fronts, but the stock dipped slightly in extended trading after seeing a rather stunning run-up over the course of the year. Since January, shares of Salesforce have risen almost 36% despite dropping around 3% after the report came out today.

While Salesforce may be on track to generate $10 billion in revenue, a lot of eyes are on the back half of the year for its annual Dreamforce conference. That’s when Salesforce is expected to announce and unveil some new products, and there will likely especially be scrutiny on its so-called “Einstein” AI products. While Salesforce birthed tools for online customer relationship management, it faces the ever-present threat of smaller companies picking off niches and turning them into big companies.

Salesforce is going to modernize its products as it goes forward, including giving companies ways to reduce workload algorithmically as machine learning starts to bleed into enterprise software. The company has been pitching that line so hard and has gone so far as to state that it’s possible to spin up customer service tools in less than a day. Salesforce is going to likely have to continue pushing this story going forward.

There are certainly a lot of opportunities here, as Salesforce has started to make a push by opening up some tools to developers. So there are going to be a lot of expectations baked into the back half of the year toward that conference as Wall Street looks to the company to unveil a new wave of products that can automate routine tasks on its services.

Wall Street was looking for earnings of 32 cents per share on revenue of $2.51 billion.

Featured Image: David Paul Morris/Bloomberg/Getty Images

News Source =

Go to Top