Turbotodd

Ruminations on tech, the digital media, and some golf thrown in for good measure.

Archive for the ‘artificial intelligence’ Category

Didi Chuxing Cha-Ching

leave a comment »

Greetings from the Big Apple.

I arrived here over the weekend to visit some friends and prepare for some meetings in NYC. 

The weather has been beyond spectacular — if I’d have planned ahead, I would have brought my golf clubs and teed up in the middle of 5th Avenue to attempt my first mile long drive.

But instead, I’m following the attempts of China’s Didi Chuxing Technology Co. to drive for a humongous IPO that The Wall Street Journal is claiming could happen as soon as this year.

Didi operates China’s largest ride-sharing platform and is expanding in Latin America and other parts of Asia, and according to the Journal report, is hoping to garner a valuation of at least $70 to $80 billion if it goes public.

The report also suggests that Didi is looking to “amass a large war chest to fend off rivals in China and other countries.”

But the company is also apparently looking to develop a smart car customized for ride-sharing and looking for auto makers that could manufacture such a car. 

The car is anticipated to be an electric vehicle and would be connected to the internet, allowing Didi to monitor data from the car for safety by applying artificial intelligence technology.

The Journal article suggests this worries some automakers, as it would put companies like Didi (and potentially others who move in this direction) in direct competition, one which could put the Didis of the world in the driver’s seat when it comes to the “operating system” for cars (i.e., the software).

Written by turbotodd

April 24, 2018 at 8:12 am

AI Funding and Talent

leave a comment »

I was too busy to blog yesterday, but a couple of stories about AI funding did hit my radar that I wanted to mention.

TechCrunch reported first that a startup out of London, BenevolventAI, announced that it had raised $115 million “to continue developer its core ‘AI brain’ as well as different arms of the company that are using it specifically to break new ground in drug development and more.”

That round values the company at $2.1 billion. 

Some background:

The core of BenevolentAI’s business is focused around what Mulvaney describes as a “brain” built by a team of scientists — some of whom are disclosed, and some of whom are not, for competitive reasons; Mulvaney said: There are 155 people working at the startup in all, with 300 projected by the end of this year. The brain has been created to ingest and compute billions of data points in specific areas such as health and material science, to help scientists better determine combinations that might finally solve persistently difficult problems in fields like medicine.

The crux of the issue in a field like drug development, for example, is that even as scientists identify the many permutations and strains of, say, a particular kind of cancer, each of these strains can mutate, and that is before you consider that each mutation might behave completely differently depending on which person develops the mutation.

This is precisely the kind of issue that AI, which is massive computational power and “learning” from previous computations, can help address. (And BenevolventAI is not the only one taking this approach. Specifically in cancer, others include Grail and Paige.AI.)

Another one that caught my attention was Eightfold.ai, “a new technology service aimed at solving nothing less than the problem of how to provide professional meaning in the modern world.”

Founded by former Googler and IBM researcher Ashutosh Garg (who is a search and personalization expert), the company “…boasts an executive team that has a combined 80 patents and more than 6,000 citations for their research.

What’s more interesting to me is their mission: “To bring the analytical rigor for which their former employers are famous to the question of how best to help employees find fulfillment in the workforce.”

Lightspeed Ventures and Foundation Capital are among those backing the venture to the tune of $24 million.

How it works:

“We have crawled the web for millions of profiles… including data from Wikipedia,” says Garg. “From there we have gotten data round how people have moved in organizations. We use all of this data to see who has performed well in an organization or not. Now what we do… we build models over this data to see who is capable of doing what.”

There are two important functions at play, according to Garg. The first is developing a talent network of a business — “the talent graph of a company,” he calls it. “On top of that we map how people have gone from one function to another in their career.”

Using those tools, Garg says Eightfold.ai’s services can predict the best path for each employee to reach their full potential.

Did you get that? “Building models for the talent graph of a company and how people have gone from one function to another in their career. I’m calling it a Maslowe AI play!

As for how hot the war for AI talent is, check out this New York Time’s article.  It reveals that AI specialists with little or no industry experience can make between $300K and $500K a year in salary and stock. 

Might be time to go back to school!

Written by turbotodd

April 20, 2018 at 12:47 pm

No Laughing Matter

leave a comment »

Did you hear the one about the personal voice assistant that, for seemingly no apparent reason whatsoever, started breaking into strange laughing noises at random?

No?

Well, I heard about it firsthand, but apparently I missed the opportunity to hear random guffawing of my own personal Amazon Tap.

According to Bloomberg, Amazon confirmed yesterday that in rare circumstances, the voice assistant can mistakenly hear the phrase “Alexa, laugh,” which under its normal programming would cause it to chuckle. 

Amazon has updated a fix for the problem, and is changing the trigger phrase for laughing to “Alexa, can you laugh?” instead.

A few moments ago, I tried the new command, and all I got from Alexa was a “Tee hee.”  

How very anti-climactic.

This quirk has been referred to in AI circles as a “false positive.”

Let’s just hope the voice commands for the AI algos running the armed drones have their laughs in order.

Written by turbotodd

March 8, 2018 at 9:02 am

Real Not Fake News

leave a comment »

Read about Farhood Manjoo’s “slow-jamming” of the news. Manjoo stopped looking at news online for two months, and instead reverted back to getting real paper newspaper subscriptions. 

I won’t give away the whole story, but one of his verdicts was that he felt as though reading an actual newspaper, he felt he was was getting some actual real news.  

What a concept, and no bots!

Of course, if bots are your thing, you may want to know about the $13.5 million Series A round that Voicera just raised.

Voicera’s elevator pitch is that it wants to make it simpler to record meetings and pull out action items automatically using AI. It will do so by recording and creating a transcript of the meeting.

It’s technology centers around “Eva,” an AI-fueled note-taking assistant. Eva’s job is to record the meeting, create the transcription, identify the important stuff, and then send out an email with the highlights to all meeting participants.

What? Eva won’t do my to dos in the process?!  What kind of AI is this?!

 

Written by turbotodd

March 7, 2018 at 12:26 pm

Meet Sarah, Buy Car

leave a comment »

The Wall Street Journal’s “CIO Journal” published an article yesterday detailing Daimler Financial Services’ efforts to explore human digital assistants (Daimler was showing off its efforts this week at the Mobile World Congress in Barcelona.)

It’s a fascinating look at the way “human” digital assistants are evolving, and probably just as important, how quickly.  

Udo Neumann is the company’s global chief information officer.

Here are a few snippets detailing Daimler’s progress:

An assistant with a human-like “face,” with instant access to helpful data and programmed to detect how people are feeling and respond accordingly, could help gain customer and employee trust, Mr. Neumann said. “It’s clearly the next step in the development of an evolving technology, (where) emotions come into play.”

Daimler Financial Services, a division of Daimler AG, announced this week it’s partnering with New Zealand startup Soul Machines on a proof-of-concept project to see how a digital assistant with a face and a name could give personalized help to employees and customers.

The companies, which have worked together for several months, are developing a “digital human” built with AI software from IBM Watson that can be programmed to answer questions related to car financing, leasing and insurance, and capabilities to recognize non-verbal cues using face recognition technology.

Neural networking and machine learning tools lets an early version, named Sarah, react to spoken and typed words as well as non-verbal queues such as a loud noise or a nodding head in agreement.

Sarah can be programmed with highly specialized knowledge about, for example, the latest Mercedes models and information about leasing options, said Greg Cross, chief business officer at Soul Machines.

The digital human could eventually act as a “companion” for employees at a call center or training center, he said. For customers, talking to such an avatar might increase purchases among those who feel intimidated by high-pressure sales staff, said Mr. Cross.

So will you one day be buying your next car from a soulful digital assistant like Sarah? 

Never say never…now this baby right over here, you just can’t go wrong. She hits 0-60 in five seconds….!

Written by turbotodd

March 2, 2018 at 10:29 am

Google’s New Machine for Learning

leave a comment »

The Verge is reporting that Google has introduced a new artificial intelligence/machine learning training website for anyone looking to learn about machine learning concepts, and develop and hone their machine learning skills.

According to the report, the site also features a free course called “Machine Learning Crash Course,” one based off an internal Google course that was originally designed to give the company’s employees a practical introduction to AI and machine learning fundamentals.

The course lasts roughly 15 hours, and includes interactive lessons, lectures from Google researchers, and over 40 exercises. The Verge reports that it is designed for newcomers with no machine learning  experience.

Written by turbotodd

March 1, 2018 at 9:11 am

Perspectives on AI

leave a comment »

​​
MIT’s "The Download" recently reported that China’s artificial intelligence startups scored more funding that America’s last year.

Of $15.2 billion invested globally in 2017 in AI, 48 percent went to China and 38 percent to America. That’s the first time China’s AI startups surpassed those in the U.S. in terms of funding.

But The Download also observes competition continues to be fierce across the board. AI startup investment rose 141 percent in 2017, and 1,100 new AI startups appeared last year.

The R&D and overall AI market may, in fact, be moving too fast.

In a separate report from Science Magazine, an analysis revealed that AI may be grappling with a replication crisis when it comes to research:

AI researchers have found it difficult to reproduce many key results, and that is leading to a new conscientiousness about research methods and publication protocols….The most basic problem is that researchers often don’t share their source code. At the AAAI meeting, Odd Erik Gundersen, a computer scientist at the Norwegian University of Science and Technology in Trondheim, reported the results of a survey of 400 algorithms presented in papers at two top AI conferences in the past few years. He found that only 6% of the presenters shared the algorithm’s code. Only a third shared the data they tested their algorithms on, and just half shared "pseudocode"—a limited summary of an algorithm. (In many cases, code is also absent from AI papers published in journals, including Science and Nature.)

Why are researchers holding back?

The article argues researchers believe some code may be a work in progress, or could be owned by a company or held tightly by a researcher eager to stay ahead of the competition.

IBM Research offered some assistance a the recent AAAI meeting, a tool for recreating unpublished source code automatically. Itself a neural network, it scans an AI research paper looking for a chart or diagram describing a neural net, parses those data into layers and connections, and generates the network in new code.

At this week’s Index | San Francisco conference, on Wednesday at 9 AM PST, New York Times journalist and author John Markoff will be hosting a session entitled "Perspectives on AI." You can register to watch the livestream here.

Written by turbotodd

February 19, 2018 at 10:11 am

Posted in 2018, AI, artificial intelligence

Tagged with , ,

%d bloggers like this: