Driverless cars are on a long and bumpy road; AI and the media: it's complicated; Big Tech's investments in AI startups mapped; US vs Europe AI divide; AI for schools
TIME profiles Anthropic; AI chatbots put to the test; Discord is where the gen AI community meets; France wants to be the new home for AI startups; internal divisions at OpenAI still linger
Many years ago, I was working for a semiconductor company that designed the multi-threaded CPU powering the Mobileye EyeQ chip. Mobileye was at the time the market leader in Level 2 driver-assistance systems, the precursor to today’s more advanced autonomous vehicle technology. The company’s product roadmap was transitioning away from camera-only processors to sensor fusion chips that used neural networks and a combination of cameras, LIDAR and other radar-type sensors to help it build a system that one day aspired to power self-driving vehicles.
Mobileye were also a supplier for Tesla and this new direction upset Elon Musk because he believed self-driving cars should only rely on cameras and nothing else, in order to keep costs down. He decided to part ways with Mobileye and set up his own chip design team that eventually produced the hardware on top of which Autopilot and then Full Self-Driving (FSD) were built. Since then, Mr Musk has spent the last decade making lofty promises on Twitter about self-driving cars being just around the corner.
We're now in 2024 and despite billions invested in the technology, we're no closer to that reality. The development of true self-driving cars that can operate without any human oversight (or what’s known as Level 5 autonomy) has proven to be an immense technical challenge that may not be solved for another decade or more.
So I wasn’t surprised to learn this week that Oxa, a startup spun out of the University of Oxford in the UK and one of the early developers of autonomous vehicle software, was forced to make layoffs, including among its executive team. Oxa’s leadership explained to departing staff that they had to make the layoffs because the company had failed to commercialize the technology quickly enough, despite raising a Series C round of $140 million in January 2023. Layoffs are incredibly tough - particularly for startups - and my heart goes out to those who were affected by them. I also want British companies like Oxa or Wayve to succeed because I don’t think we should live in a world where critical parts of our transportation infrastructure are developed only in the United States or China. And as I think about the current crop of companies developing autonomous driving technology, Oxa had the most reasonable strategy of them all: rather than going after ride-hailing consumer apps, they targeted more industrial-focused applications where the technology could be deployed in controlled environments such as passenger shuttles in airports or goods deliveries.
But inevitably we need to be realistic about the limitations of self-driving vehicles. Hardware is hard. Unlike the digital world where Google’s AI hallucinating about glue as a pizza topping is unlikely to cause immediate harm, autonomous vehicles malfunctioning can have life-ending consequences. Driving in the real world involves an incredible number of changing variables, edge cases, and split-second decisions that are remarkably difficult for even the most advanced AI systems to handle safely at this point. From reliably recognizing and responding to jaywalking pedestrians to navigating through construction zones to handling severe weather conditions, there are countless potential hazards that self-driving software must be able to flawlessly interpret and react to without any room for error.
Major Level 4 self-driving programs like the ones launched by Waymo and Cruise in the United States have been able to deploy limited, inner-city ride-hailing services with backup human safety drivers monitoring the vehicles. But removing the human safety net completely and allowing broad self-driving vehicle deployment requires a standard of AI perception, prediction, and control that still remains years away, based on the current rate of progress.
As someone who loathes everything about the experience of driving in the UK (the endless traffic jams and road closures, the bad drivers, the potholes, the shortage of parking), I can’t wait for self-driving cars to become a reality. But while I daydream of taking a nap while I’m being driven to the St Albans train station so I can start another soul-crushing commute, the technical realities mean we're still a decade or more away from that full autonomy becoming a reality outside of small, controlled areas. Until self-driving vehicles can become truly infallible even in the most chaotic and unexpected driving scenarios, automakers and regulators will rightly remain cautious about unleashing them en masse.
So to wrap up, I want to take you back to a nondescript conference room on the sidelines of CES 2014 in Las Vegas where I saw prof. Amnon Shashua, the CEO and president of Mobileye, for the first time. He had a slide up on the screen showing “autonomous driving megatrends“ together with his predictions for the “near future” and “future.” During Q&A, he said one thing that stuck with me, and which I believe still remains true today: it is easy to get an autonomous system from 0 to 95 percent but no one has yet figured out how to solve the last 5 percent. Until that happens, the road ahead for self-driving cars remains long and bumpy.
And now, here are this week’s news:
❤️Computer loves
Our top news picks for the week - your essential reading from the world of AI
Fortune: How Waymo outlasted the competition and made robo-taxis a real business
Washington Post: The media bosses fighting back against AI — and the ones cutting deals
Time: Inside Anthropic, the AI Company Betting That Safety Can Be a Winning Strategy
FT: Internal divisions linger at OpenAI after November’s attempted coup
TechCrunch: Big tech companies are plowing money into AI startups, which could help them dodge antitrust concerns
The Information: OpenAI CEO Cements Control as He Secures Apple Deal
WSJ: The Great AI Challenge: We Test Five Top Bots on Useful, Everyday Skills
Wired: Does String Theory Actually Describe the World? AI May Be Able to Tell
Axios: Axios interview: Sonos CEO talks AI, Apple and consumer electronics
Bloomberg: France Envisions AI Tent Big Enough for US, China and French Startups
AFP: OpenAI's Johansson gaffe pushes voice cloning into spotlight
Business Insider: Inside Amazon's struggle to crack Nvidia's AI-chip dominance
TechCrunch: Discord has become an unlikely center for the generative AI boom
FT: AI for Schools
⚙️Computer does
AI in the wild: how artificial intelligence is used across industry, from the internet, social media, and retail to transportation, healthcare, banking, and more
Axios: Moxie is an AI robot offering kids a supportive companion
The Verge: Alexa’s Fire TV search has a new AI, but it needs some work
TechCrunch: Autobiographer’s app uses AI to help you tell your life story
MIT Technology Review: AI-directed drones could help find lost hikers faster
The Verge: How AI could change EV charging
TechCrunch: Microsoft’s Copilot is now on Telegram
The Verge: YouTube Music will let you search by humming into your Android phone
Business Insider: Liverpool FC's content strategy uses AI to bring the action on the field to fans' phones quickly
The Verge: Chromebook Plus laptops are getting Google Gemini
🧑🎓Computer learns
Interesting trends and developments from various AI fields, companies and people
MIT Technology Review: Why are Google’s AI Overviews results so bad?
VentureBeat: McKinsey: Gen AI adoption rockets, generates value for enterprises
Axios: AI data center demand poised to juice U.S. power needs
VentureBeat: Anthropic's Claude AI now autonomously interacts with external data and tools
Bloomberg: Apple Plans AI-Based Siri Overhaul to Control Individual App Functions
Bloomberg: OpenAI CTO Says Generative AI’s Economic Impact Only Starting
VentureBeat: Perplexity goes beyond AI search, launches publishing platform 'Pages'
CNBC: A top secret meeting is kicking off in Madrid with AI CEOs leading the attendees
VentureBeat: You.com empowers users to create personalized AI assistants with top language models
WSJ: Klarna Marketing Chief Says AI Is Helping It Become ‘Brutally Efficient’
Reuters: Big Tech develops AI networking standard but without chip leader Nvidia
Reuters: OpenAI offers nonprofits discounts on corporate ChatGPT product
The Atlantic: A Devil’s Bargain With OpenAI
Fortune: Gen AI at work has surged 66% in the U.K., but bosses aren’t behind it
VentureBeat: Mistral announces Codestral, its first programming focused AI model
TechCrunch: AI models have favorite numbers, because they think they’re people
New York Times: How A.I. Made Mark Zuckerberg Popular Again in Silicon Valley
Axios: Sam Altman is starting to look a lot like Mark Zuckerberg
Semafor: France is an AI hub, but a wrinkle in tax policy is holding it back
The Information: Apple’s Plan to Protect Privacy With AI: Putting Cloud Data in a Black Box
VentureBeat: SambaNova breaks Llama 3 speed record with 1,000 tokens per second
VentureBeat: SAP adds Amazon Bedrock into AI Core, streamlining generative AI use for regulated firms
Bloomberg: Nobel Laureate Romer Says AI Hype Risks Repeat of Crypto Bubble
Business Insider: PwC inks deal with OpenAI for powerful ChatGPT version as almost all its consulting clients 'actively' engage with AI
Reuters: Arm offers new designs, software for AI on smartphones
Fortune: Why SAP’s CTO made scale a strategy for generative AI
The Economist: Can Elon Musk’s xAI take on OpenAI?
Business Insider: What productivity boom? AI will provide just a 1% GDP boost over the next decade, MIT economist says.
Reuters: OpenAI signs content deals with The Atlantic and Vox Media
Business Insider: Creators are turning away from some AI tools, new data shows+
Business Insider: How AI and virtual influencers are invading YouTube and social feeds
VentureBeat: Microsoft, Beihang release MoRA, an efficient LLM fine-tuning technique
Business Insider: OpenAI CEO Sam Altman pledges to give away the majority of his wealth
TechCrunch: Skej’s AI meeting scheduling assistant works like adding an EA to your email
Bloomberg: Mistral AI, France’s Startup Darling, Takes Aim at the US Market
Washington Post: Apple and Microsoft want you to notice how their AI works. But why?
New York Times: If A.I. Can Do Your Job, Maybe It Can Also Replace Your C.E.O.
Bloomberg: China Chipmakers Catching Up Fast in AI, SenseTime’s Xu Says
Business Insider: OpenAI's former executive Jan Leike joins its rival Anthropic days after quitting over safety concerns
Bloomberg: Ex-DeepMind Trio Bring Algos, AI Poker Prowess to Tower Research
Business Insider: Netflix chief doesn't think AI programs are going to replace writers, actors, and directors
TechCrunch: Iyo thinks its gen AI earbuds can succeed where Humane and Rabbit stumbled
The Information: Musk Plans xAI Supercomputer, Dubbed ‘Gigafactory of Compute’
TechCrunch: Doly lets you generate 3D product videos from your iPhone
WSJ: Should AI Have Access to Your Medical Records? What if It Can Save Many Lives?
FT: Meta and Elon Musk’s xAI fight to partner with chatbot group Character.ai
The Telegraph: Director creates 50,000 AI ‘actors’ to hire for crowd scenes
Business Insider: Apple may soon bring generative AI to emojis. Here's what that means.
WSJ: The Student Who Was Suspended for a Prizewinning AI Tool Fights Back
Business Insider: A Gen Z software engineer studied AI in college. He says you shouldn't major in it if you fall into one of these 3 categories.
Business Insider: Educating workplaces about AI will lead to better outcomes, say business leaders
Business Insider: AI could take investor relations to the next level and 'deliver more value,' says a tech CEO
The Guardian: Trying to tame AI: Seoul summit flags hurdles to regulation
Keep reading with a 7-day free trial
Subscribe to Computerspeak by Alexandru Voica to keep reading this post and get 7 days of free access to the full post archives.