New rules for testing and deployment of self-driving vehicles are on the way in the state of California.
The California DMV announced on Friday that it is now seeking public comment on proposed regulations that would apply to self-driving vehicles from light-duty robotaxis like Waymo’s to heavy-duty driverless trucks like those from Plus.AI.
The industry’s 2010s-era foray into self-driving didn’t really pay off for many automakers. General Motors-backed Cruise lit money on fire while others, like Ford’s Argo AI, died in infancy. And while Tesla has steadily improved its Autopilot and Full Self-Driving features over the past decade and even launched a Robotaxi service, it too hasn’t achieved its biggest promises.
But things seem to be changing thanks to the boom in AI and computing power. Sterling Anderson, former co-founder of self-driving startup Aurora and former head of Autopilot at Tesla, joined GM earlier this year to boost its entire tech game—including autonomy.
Tesla has the more accurate business model of Integrating automated self-driving with their vehicles. Once the technology is capable of handling all circumstances and weather conditions it’ll explode almost overnight. Likely with the completion and integration of the AI data centers is when we’ll see this technology really take off.
need time frame. otherwise, sure in 100 years
same with “once China has global naval superiority they’ll invade”. Need time frame for useful discussion
Haha, fair point.
My estimation is a 6 to 10 year timeframe. The AI Datacenters that are part of Amazon Web Centers (AWC), Project Stargate, and Project Rainier will have completed most of their construction and stand-up implementation by the end of 2025. So 2026 to 2036 will see enormous increases in AI infrastructure and automation functions implemented into daily life as networks become more dense and integrated into hand-held technology and urban cities.
What I expect to see:
- Automation and fine-tuning of electronic traffic control devices with AI algorithm and adjustment.
- Handheld devices will all come with a subscription or built-in AI assistant service that will replace or displace many apps
- Financial institutions will automate more of their stock market and trading functions to AI. Faster speeds for stock trades = more money for the company that pays the most.
- As AI develops more thorough integration into daily life the concept of “Google Glasses” and customizable AI assistants will again take off and have even greater impact as human-technology collaboration assistants
- Vehicles will have develop hybrid technology for self-driving by individuals and automated driving run and operated by AI. Insurance companies will incentivize AI driven vehicles as AI vehicles will have an artificially low accident rate compared to human-operated vehicles.
When for the hovercraft delorean time machine?
Well it ain’t ever taking off here with the taxi gangs getting anything they want with the click of a finger.
Not that I would want it too but anyway, for Taiwan… its a moot point.
Unfortunately my outlook is a bit more… pessimistic.
I think we’re more likely to see AI for nefarious internet use cases, rather than robotic automation in daily life:
-
OpenAI said they are pivoting into porn and advertisements because it’s more profitable (Bloomberg).
-
Social media (IG, FB, TikTok) will have “infinite scroll” of AI content to more effectively rot your kid’s brain
-
More convincing scams with real-time deepfake video calls operated by humans
-
More AI services to cheat during job interviews, immigration visas, and school homework
-
Websites/apps become more difficult to use, as banks/businesses require more layers of authentication due to deepfakes
On the bright side, influencer culture will probably die.
Robotics has always inspired the general public and given hope about a brighter, better future. But robotics never took off in any of the AI summers (expert systems in 1980s, big data in 2000s, deep learning in 2010s, and now LLMs in 2020s).
The reason we haven’t seen this automated future is primarily because of social and government roadblocks. We had the technology for self-driving back in the 2004 (DARPA Challenge).
If governments cared enough, they could build highways restricted to only self-driving cars. This could easily automate an important part of supply chain logistics, without the decades of R&D required for city driving.
This doesn’t need LLMs, and doesn’t even need linear regression. Traffic engineers designed roads and traffic lights for decades without LLMs.
Even 10 years ago, smartphones sent your GPS (and acceleration, compass heading, …) every second to the cloud. If cities cared enough, they could pass legislation to require data sharing for optimal traffic light control to minimize commute times and vehicle idling in areas prone to bad air quality. But the juice is not worth the squeeze. If you reduce traffic, new people will come and restore the original traffic (see Los Angeles).
That’s what they said about the smartphone, but we ended up getting an app for everything. We continue to see that trend with LLMs/chatbots. Every website, every company has their own damn chatbot. Companies rarely hand over their data to someone else. And today, a CEO will get a larger bonus if they say “we built our own LLM” rather than saying “we are giving our data to OpenAI to build a unified super intelligence to improve the lives of us all”.
At best, you’ll have an AI assistant (technical term is “agent”) that will open some websites/apps (e.g. Agoda) for you and do some of the tasks. But keep in mind, if it works too well, Agoda will revoke Google/Apple’s agent access, and Agoda will build their own agent because it’s cheaper. And CEO gets a bonus.
This one definitely not.
Modern AI/LLMs are extremely slow. The best high frequency trading firms respond to news/market events in nanoseconds. DeepSeek was created by a hedge fund who does not compete on speed (Chinese stock markets do not provide real-time information to traders, unlike the rest of the world).
Simply moving data from GPU to the network card (so you can interact with the internet/stock exchange) requires microseconds, and this is after the GPU ran the AI computation, which itself requires seconds.
In 100 nanoseconds, the speed of light covers 30 meters. In 1 second, it travels to the moon.
This is more of a social/privacy acceptance problem. People don’t like being recorded, especially by government.
Second problem is battery life of these wearables.
You can run AI models locally on-device or broadcast to the cloud via 5G cell towers. In both cases, battery consumption is too high. Snapchat’s Spectacles AR glasses last less than 1 hour, and that’s just video recording. No AI processing.
Once that happens, yes.
I believe Tesla Insurance is the only one offering discounts if your car has FSD? And it’s mostly a marketing stunt to improve confidence in Tesla’s FSD and pump the stock price. Tesla’s FSD accident rate is incredibly high when off the highway on city streets/traffic lights. That’s why Tesla always reports their accidents per million miles including highway miles (where nothing happens).
Keep in mind, we don’t have self driving semi-trucks for highways, and that’s an easier problem than self driving cars in cities. You can easily have a truck driver (new job title will be security guard to protect from pirates) drive the last-mile from highway off-ramp to the distribution center.
But overall, in my opinion, universal basic income (UBI) is required before you see significant AI automation in daily life.
The impact on humans/jobs is immense. So until UBI is being seriously considered by politicians, robotic automation is very far away (20+ years).
Classic movie, but no dice. 
As soon as Taiwan has 50 000 drones their naval superiority is obsolete.
The number depends on what kind of military grade drones we are considering though
Both can be true.
I’m not surprised at the pivot towards porn and advertising. Blu-Ray won out over HDDVD as a medium because of the porn industry. All the above I agree are likely to become true along with declining personal privacy protections.
Road construction is an expensive endeavor to build and maintain especially when asphalt surface roads are the dominant material used for construction. Concrete would require less maintenance over the same period of time, but it’s not as easily repaired and if replacement is needed the up-front costs tend to be far more prohibitive.
https://www.ardot.gov/wp-content/uploads/2021/11/2020-CPM.pdf
That’s the current state of the software environment, but the current chatbots and LLM’s will be dwarfed by the AI capabilities that may be introduced within the next decade. It would be equivalent of the jump from mechanical windows all car’s had in the 90’s to todays Tesla that has limited self-driving capabilities. It’ll develop slowly, and suddenly then be everywhere all at once.
The heavy lifting would have to be done by a smartphone. The glasses would transmit data, but store nothing on the device itself. All of it would be tied to your phone.
Robotic automation is slowly taking over repetitive tasks and menial jobs so I don’t see it being that far away. The current cost of designing and building robotics devices is slowly trending down though. Initially it will be “robots for rent” to get companies onboard with using more robotics with less capital risk and expenditure. As newer and better versions are developed and produced the ubiquitousness will grow more rapidly than government is able to respond to the changes that follow. UBI is not something I expect to see within 30 to 40 years, and by then the world will be very different compared to today.