National

Whats Next

Axios What's Next

Cruise and Waymo are plowing ahead with plans to bring robotaxis to more cities, despite growing disillusionment among investors and automakers about the timeline for self-driving cars.

Why it matters: The two companies already operate driverless taxis in San Francisco and Phoenix. 

If they can exploit that early lead by quickly scaling ride services across multiple cities, they have a chance to prove the economics of self-driving technology are viable and that it's not just an expensive science experiment. 

Catch up quick: Other key players are shutting down, merging or paring back.

Ford and Volkswagen last month pulled the plug on their autonomous vehicle joint venture, Argo AI. 

AV delivery startup Nuro laid off 20% of its workforce, and two leading AV tech suppliers announced a merger.

Even Waymo's parent, Alphabet, is being pressured by an activist investor to reduce losses at its self-driving unit.

What they're saying: "Profitable, fully autonomous vehicles at scale are a long way off," says Ford CEO Jim Farley.

What's happening: Far from retreating, Waymo and General Motors-backed Cruise are hitting the accelerator.

"We're going to continue to make sure we don't constrain the commercialization because we have the lead right now," GM CEO Mary Barra told investors last week. 

"Those who are writing that it's not going to work and that it's decades off haven't taken a ride in the vehicle — I mean, we're doing it right now."

What's next: After San Francisco, Cruise plans to expand robotaxi service to Phoenix and Austin, Texas, by the end of December.

Several more cities are planned for 2023, and Cruise CEO Kyle Vogt said the company aims to hit $1 billion in revenue by 2025.

Waymo is expanding too, announcing that Los Angeles will be its third market (after Phoenix and San Francisco).

Last Friday, California gave Waymo the go-ahead to transport passengers in San Francisco without anyone behind the wheel, although it still needs a separate permit to begin charging fares as it does in Arizona.

Waymo is also unveiling a dedicated robotaxi co-developed with Zeekr, an electric vehicle brand owned by China's Geely Automobile Holdings.  

Meanwhile, another AV company, Motional, is partnering with Uber and Lyft to add Hyundai robotaxis to their ride-hailing fleets in several cities starting next year.

The big picture: This could be the beginning of an industry-wide shakeout in which the leaders pull away from the pack, similar to what happened in the auto industry in the first half of the 20th century.

Back then, hundreds of automobile companies fell by the wayside, leaving only three: GM, Ford and Chrysler.

Or, it could simply be that — unlike some others — Waymo and Cruise haven't run out of money yet. 

One big advantage Cruise has is a $5 billion line of credit from GM Financial, the automaker's lending arm, to fund expansion.

Still, the pressure is on for both companies to prove they can make money on self-driving technology — and the only way to do that is by scaling up quickly. 

Cruise's Vogt is bullish, despite what others are saying. 

"If you look at the underlying facts, you have people actually using driverless cars in major cities for the first time, multiple announcements of expansion, increasing fleet size, increasing metrics," he said in an interview. 

The bottom line: Vogt believes "we are in the golden years of AV expansion."

"It's right in front of everyone," he tells Axios. "But I think it's one of those things that won't really be fully comprehended until a year or two from now — when AVs are practically commonplace and people are looking around and saying, 'Where did this come from?'" 

Cruise, Waymo push robotaxis amid doubts about self-driving tech

Robotaxis are making inroads, but will they have staying power?

Waymo plans to deploy robotaxis co-developed with Zeekr, an electric vehicle brand owned by China's Geely Automobile Holdings. Photo courtesy of Waymo

Autonomous vehicles

Automation and AI

Transportation

From the Valley to D.C., Big Tech players like Facebook, Google and Amazon are under more scrutiny than ever as new technology develops and privacy and antitrust concerns grow in lockstep with companies’ ambitions.

Technology

After a record economic expansion, the coronavirus pandemic sent the U.S. economy into a recession as unemployment soared to staggering heights. The country now faces urgent questions about how much stimulus is needed for reeling consumers and businesses, and what a recovery might look like.

Economy & Business

Login

Axios Login

AI's magical text-to-image generators, like Dall-E 2, have sparked fears of unemployment among professional illustrators — but Adobe, the leading maker of software tools for designers, sees AI as more of a creative assistant.

Driving the news: At at a conference this week, Adobe showed how it could build generative AI tools into Photoshop, Lightroom and other products.

Why it matters: Fanciful images generated by the likes of Dall-E 2 and Stable Diffusion have raised thorny legal and ethical questions, but Adobe's early work provides a glimpse at an alternative vision of humans and AI working together.

Details: At its MAX conference in Los Angeles this week, Adobe showed off a number of ways generative AI could help creative workers without entirely supplanting artists.

In Photoshop, Adobe showed how AI could offer artists an early look at a number of different creative approaches to a subject, allowing them to quickly explore options in far less time.

In Lightroom, Adobe showed how generative AI could be used to enable a slider that allowed the background sky of an image to be changed gradually from day to night. (Adobe already has AI-infused neural filters that do specific transformations.)

With its newer web-based Express tool, Adobe showed how a text prompt could be used to transform any of its thousands of fonts to incorporate other objects. In one case, Adobe merged fonts with electrical cables, and in another it infused the type with orchid petals.

The big picture: Adobe's approach offers a contrast to the projects put out by OpenAI, Google, Meta, Stable Diffusion and others, which have largely focused on what AI programs can do on their own given text prompts from users. 

What they're saying: While some see generative AI as a threat to artists' livelihood, Adobe chief product officer Scott Belsky told Axios he sees it as eliminating mundane chores, similar to the way GitHub's Copilot helps  programmers code faster.

"It helps you achieve more explorations in less time," Belsky said in an interview Thursday.

Belsky also sees an opportunity for the software to help people know which elements of an image came from an artist and which used AI (and perhaps which AI system), drawing on work from the company's deepfake-fighting Content Authenticity Initiative. "People need to know where this stuff is coming from," he said.

Yes, but: Legal issues remain, particularly around intellectual property and rights to the data used to train these AI programs. Belsky said that Adobe is wrestling with the same questions as others in the industry, and also trying to develop business models that support artists.

One approach Adobe is weighing: rely on a so-called clean model that is trained only on images to which Adobe has full rights. 

Another path could be to allow creators in Adobe's Behance network to opt in, allowing their work to be used and potentially collect royalties when another user creates work in their style using the AI tools.

Then there is the question of who owns an AI-generated image. 

While U.S. law doesn't allow fully computer-generated works to be copyrighted, Belsky said he is confident that artists will be able to incorporate the output of generative AI into their own work without sacrificing their rights of ownership. 

What's next: While much of the industry's early work has focused on image generation (and simple videos), Adobe sees generative AI also being useful in everything from 3D design to texture creation to making logos.

Adobe sees AI helping artists, not eating their lunch

The software giant plans to build generative AI into its creative products as an aid for designers and artists.

An Adobe Express demo using generative AI and fonts

Artificial intelligence

Column / Signal Boost

Google is touting its role in helping transform Ford Motor Company as an example of what other large businesses can gain by making use of its cloud services.

Driving the news: Ford CEO Jim Farley will join Google Cloud chief Thomas Kurian today during Google Cloud Next to offer an update on how the two companies' relationship is evolving 18 months into a six year partnership.

Google is helping Ford with a range of tasks from quality assurance in its factories to delivering over-the-air software updates to vehicles.

The deal also includes a separate commitment by Ford to use Google's Android operating system to power the in-car entertainment systems of future vehicles.

Why it matters: Despite their size, both companies have their work cut out.

Google's cloud business has been fast-growing, but still significantly trails market leaders Amazon and Microsoft.

Ford, like other traditional auto makers, is trying to secure its spot in a significantly changing world in which electric engines are replacing gas power and technology plays a key role in everything from music to self-driving features.

What they're saying: “We’re proud to work with Ford and apply the best of Google’s AI, data analytics, compute and cloud platforms to help it deliver pioneering work as it transforms from an automaker into a software company that also sells cars," Kurian said in a statement to Axios.

Exclusive: Google touts cloud deal with Ford

Cloud services are a big part of how Ford is reinventing itself, according to Google.

Google CEO Thomas Kurian, speaking at the Google Cloud Next '19 event in San Francisco, California

Google Cloud CEO Thomas Kurian, speaking at the Google Cloud Next '19 event in San Francisco. Photo: Michael Short/Bloomberg via Getty Images

Google

Facebook parent company Meta Thursday announced Make a Video, an online tool that generates short movie clips based on a text description.

Why it matters: Text-to-still-image AI systems, including DALL-E 2, Stable Diffusion, and rival projects from Google, Meta and others, have advanced with great speed over the past year. Now, Meta appears to have won the race to extend AI-based content creation to videos.

Details: As with the still-image generators, all people have to do is describe something they want to see in a verbal prompt and the system returns a visual result — in this case, a video.

Make a Video can also create videos from images or take existing videos and create new ones that are similar, Meta said.

Meta is not yet making the tool available for public testing but plans to do so by November.

The big picture: AI image generators have wowed observers, but they also raise significant concerns regarding ownership, privacy, the data sets they're based on and the impact their advent may have on artists' income and employment.

What they're saying: "We want to be thoughtful about how we build new generative AI systems like this," Meta said in a blog post. "We are openly sharing this generative AI research and results with the community for their feedback, and will continue to use our responsible AI framework to refine and evolve our approach to this emerging technology."

Go deeper: The Washington Post tells the story of research work at OpenAI that pioneered much of this new field.

Meta announces AI-based tool for generating video from text

The tool builds on popular image-generation programs like DALL-E and appears to have won a race to extend the tech to video.

A thumbnail of videos created using Meta AI's Make a Video tool

Thumbnail images of videos generated by Meta AI's "Make a video" project. Image: Meta

Meta (company)

If you've ever swerved to avoid an obstacle on the highway, you'll understand why Aurora Innovation is teaching its self-driving trucks to be prepared for anything — including shredded tires or errant sofas.

Why it matters: Road debris is fairly common and can be dangerous, especially when vehicles are moving at highway speeds. 

Nearly 37% of all deaths in road debris collisions result from the driver swerving to avoid hitting an object, according to a study by the AAA Foundation for Traffic Safety. 

What's happening: The latest beta version of Aurora's self-driving truck tech, called Aurora Driver, is trained to identify and steer clear of debris on the road to avoid swerving collisions at highway speeds.

It can also navigate construction zones with temporary lane markings and identify vulnerable road users, such as cyclists and motorcyclists.

The bottom line: Autonomous vehicles have to share the road with other cars, cyclists and pedestrians, but also things that shouldn't be there — like a piece of furniture that fell off the back of a truck.

What to watch: Aurora aims to launch fully driverless semi trucks by the end of 2024. 

Self-driving trucks are learning to circumvent major road debris

Aurora's self-driving trucks learn to navigate highway debris to avoid swerving collisions

Illustration of torn and split sofa with highlighter texture

Autonomous vehicle technology

Finish Line

Axios Finish Line

Arguably the most consequential technological transformation of our lifetimes is coming faster than we think — and we're not adequately prepared.

Why it matters: Artificial intelligence is rapidly improving at human-like tasks, like language and reasoning.

And the number of plugged-in experts who say we have years and years to prepare ourselves for AI’s rise is dwindling, the New York Times lays out in a stunning new report.

Case in point: Five years ago, AI's big win was that AlphaGo — a machine built by Google — beat humans at the game Go, which was one of the last games where humans still had the upper hand.

But even that seems minor compared with last year when Google's AI solved a molecular biology problem that had confounded scientists for decades, the Times reports.

What’s happening: We asked the top experts in our newsroom exactly what AI can and can’t do — and how we should be thinking smartly about it.

Yes, AI is getting smarter. It will keep getting astonishingly better at suggesting possibilities, like videos to watch, and extending patterns, like next things to say in a chat, Axios managing editor of technology Scott Rosenberg tells us.

But for the foreseeable future, people will remain the arbiters of which possibility is actually best, or which pattern is useful. 

In this sense, AI will always be a tool — but one so powerful and able to operate independently enough that we can be forgiven for mistakenly seeing it as autonomous.

But AI's results are only as good as the algorithms and the data used to train the systems, both of which are susceptible to bias, Axios chief tech correspondent Ina Fried notes.

That helps explain, for example, why chatbots may quickly devolve into racist speech. Or why AI-generation pictures of a doctor are likely to show a white man.

🧠 Reality check: AI is far from human. Despite high-profile claims to the contrary, the machine-learning-based AIs we know today can't achieve sentience because they don't have bodies and senses to create a sense of self in time and place, Scott says.

That could change! But no one knows how to make it happen today, or even has a good theory for how to get there. And before we go down that road, we'd better have a long conversation about whether it's even a good idea.

Demystifying AI

Artificial intelligence is rapidly improving at eerily human-like tasks.

Illustration of robot hands holding a clipboard.

Lyft is adding a new fleet of electric, autonomous vehicles (AVs) to its ride-hailing network in Las Vegas, bringing it one step closer to offering a full robotaxi service.

Why it matters: It's the latest milestone in the partnership between Lyft and Motional, one of a half-dozen leading autonomous tech companies. 

Rather than designing its own self-driving cars, Lyft is partnering with others to bring AVs to its network. 

Details: Motional's new AV is based on Hyundai's electric Ioniq 5, customized for driverless ride-hail operation.

It's equipped with more than 30 sensors — cameras, radar and lidar — and an onboard computer system that helps the vehicle see and understand the world around it. 

Riders can unlock the doors via the Lyft app and start the ride using an intuitive passenger display in the rear seat. 

If needed, they can also use the screen to contact remote agents.

Of note: For now, two vehicle operators will be present in the front seat, but passengers can get a feel for what the robotaxi experience will be like. 

What's next: The service is set to become fully driverless in 2023, and will then scale to other cities.

In Las Vegas, your Lyft driver could soon be a robot

The service will eventually expand to other cities

A woman prepares to board an autonomous Lyft robotaxi in Las Vegas

Lyft and Motional are developing robotaxi service in Las Vegas. Photo courtesy of Motional

Closer

Axios Closer

Artificial intelligence has seeped into many creative trades — from urban planning to translations to painting. The latest: visualizations in journalism.

Why it matters: Computers are getting better at doing what humans can do, including creating art from scratch. But design is much more than mashing together some known shapes on top of a color background.

Artists interpret and convey abstract stories through a single image that not only has to grab someone's attention, but also stand the test of time.

Case in point: Brian Rumao, managing director at Next Play Ventures, posted a sample on Twitter using an AI image generation tool from Open AI called Dall-E 2 to try to reproduce a couple Axios visuals, to some success (see sample below).

The image on the center-left was created after inputting "a long black chain-link fence with barbed-wire dollar-signs on top, against a yellow background" into the tool, Rumao noted.

Screenshot of two tweets featuring Axios Visuals beside similar Dall-e composites 

The post generated a lot of internal discussion among the 23 visual journalists at Axios.

This led to the question: What is the difference between an editorial illustration developed by a human being and a visual composite generated by AI?

The big picture: Digital art generated from new AI systems like Midjourney and OpenAI's Dall-E 2 have prompted a lot of newsroom discussions, especially as the Dall-E 2 website explicitly mentions using the tool to generate newsletter art.

On the one hand: Machine-learning platforms can create photo composites from inputs to mimic fine art, photography or illustration. 

On the other hand: Editorial illustration requires a level of conceptualization, visual metaphor and story connection that these kinds of tools currently can't replicate. 

How it works: At Axios, illustrators begin by pinpointing key ideas from a story.

Next, they conceptualize visual representations of the themes.

Then they brainstorm how to tie the themes together to create a metaphor.

After producing several mockups, illustrators often select a version that's quickest to read — or one that's "reader first" — that conveys an idea at first glance.

An example: A story about two health care startups merging might prompt an illustrator to depict "digital," "health" and "merger" — the key elements of the story — using symbols that represent each. 

The final version might be two syringes tied together with an ethernet cable.

One issue is that AI technology can be prone to encoded biases. While people also are influenced by their biases, they can take steps to overcome these.

Editorial journalists are trained to consider these issues, and by encouraging conversations inside a diverse newsroom, they can see possible problematic interpretations of a visual.

People can also follow the evolution of language closely, and how those changes should be reflected visually. They can watch for tropes and guard against overused symbolism.

What to watch: For now, AI like Dall-E 2 is incapable of replicating the creative problem-solving process required to create successful editorial illustrations.

They are, however, turning out to be powerful tools to generate colors, ideas, patterns, textures, or mood boards for larger projects.

AI-generated digital art spurs debate about news illustrations

AI digital art may be improving — but hasn't replicated the creative problem-solving process needed for newsrooms.

Illustration of a artist's pallet with drops of paint in the shape of binary code 

GM's hands-free driving tech is set for a massive expansion

GM's "Super Cruise" is coming to 400,000 miles of North American roadways

A woman driving hands-free in a Cadillac Escalade

A Cadillac Escalade equipped with GM's hands-free Super Cruise technology. Photo courtesy of GM

Grindr is partnering with Spectrum Labs, tapping the startup's AI-based system to help filter postings on the LGBTQ dating service.

Between the lines: For years, Grindr has chosen not to implement an AI system for content moderation, not because it didn't want to augment its keyword-based filtering system, but because it was concerned that the models weren't sensitive enough to keep users safe without introducing other types of bias. 

Content moderation via machine learning is tricky, controversial and not always good, Grindr spokesman Patrick Lenihan told Axios.

With Spectrum, which provides content moderation to other dating services as well as game companies and other Internet companies, Lenihan said Grindr finally found an option it was comfortable with. "They had the thing we really needed."

How it works: Rather than simply police content for certain words or phrases, Spectrum's contextual AI service works to solve specific issues, such as identifying the sale of drugs and sex as well as trying to detect underage users.

Spectrum has a set of algorithms it has tuned over the years, but also works with each customer to make the system work for their environment. As a result, it can take weeks or months to get its tools up and running, but Spectrum CEO Justin Davis says that's an investment that pays dividends over time.

Why it matters: While Grindr had understandable reasons for waiting to find a suitable AI system, not using one meant the company was heavily reliant on user reports. In addition to being reactive rather than proactive, the approach is also vulnerable to abuse.

Spectrum's Davis says that only 18% of users across services report problematic encounters and a huge percentage of those are actually false reports, such as people who didn't like their date.

And the other non-AI method that Grindr and others use — monitoring for keywords — has gotten less effective over time as people have become more sophisticated in avoiding such systems.

The big picture: Dating apps have become the key method for matchmaking, but the rise in popularity has also made them a hotbed for harassment, illegal activity and scams.

Fraudsters bilked online daters out of an estimated $500 million last year, according to one study based on data reported to the FTC.

Exclusive: Grindr partners with Spectrum Labs on AI moderation

The LGBTQ dating site had relied on user reports and keyword filters

An illustration of a heart with an Ethernet cord wrapped around it

Dating apps

Future

Axios Future

A new report lays out the ways that cutting-edge text-generating AI models could be used to aid disinformation campaigns.

Why it matters: In the wrong hands text-generating systems could be used to scale up state-sponsored disinformation efforts — and humans would struggle to know when they're being lied to.

How it works: Text-generating models like OpenAI's leading GPT-3 are trained on vast volumes of internet data, and learn to write eerily life-like text off human prompts.

In their new report released this morning, researchers from Georgetown's Center for Security and Emerging Technology (CSET) examined how GPT-3 might be used to turbocharge disinformation campaigns like the one carried out by Russia's Internet Research Agency (IRA) during the 2016 election.

The disinformation threat from text-generating AI

A new report raises alarms over the power of systems like GPT-3 to generate vast quantities of deceptive content.

Automation and AI

The big picture

The disinformation threat from text-generating AI

AI is industrializing

How the automation economy can turn human workers into robots

AI and automation are creating a hybrid workforce

The next wave of job automation will be virtual

Automation is 2020's least understood issue

All Automation and AI stories