Apple dials back self-driving car goals to launch in 2026 • The Register

In brief Apple’s plans for an autonomous vehicle are being pushed back at least one year to 2026 after Cupertino reportedly scraped its previous designs. 

The (kind of) top secret project, known internally as Titan, is an ambitious undertaking even for the world’s most valuable company. Like many others, Apple also wants to be the first to develop a fully autonomous Level 5 self-driving car – requiring no assistance from human drivers, including pedals and steering.

It seems the technology isn’t quite there yet, forcing the company to rethink its design. Apple will now scale back initial plans and add a steering wheel and pedals. Bloomberg first reported. The autonomous car won’t have Level 5 capabilities, and will require drivers to take over in some instances.

Apple has spent years and billions of dollars trying to build an autonomous car, and plans to sell the vehicle to Joe Public for $100,000. If the company is going to start seeing some returns from the project, it’ll have to release something soon.

Simulating a virtual machine using ChatGPT

Netizens continue to be fascinated with OpenAI’s latest language model ChatGPT, posting all the weird and wacky conversations they have had with it.

One of the most bonkers examples we have seen so far is getting ChatGPT to behave as if it was a Linux terminal. Jonas Degrave, a research scientist at DeepMind, posted snippets of the text-generating model replying to prompts that coders would usually write to interact with the machine using the command line.

ChatGPT was able to generate the outputs that a Linux terminal would display for commands to search files, and even create new ones that could be accessed later on in the conversation. The model was even able to act as if it had correctly run scripts to perform simple arithmetic problems, like adding the first ten prime numbers.

It can also apparently mimic what a virtual machine would spit out if a user had asked it to create a docker file, spin up a GPU and access websites on the internet. ChatGPT doesn’t actually have the software to really carry out any of these tasks, it just simulates what the outputs of a Linux terminal would look like if used in those ways.

The eeriest thing was ChatGPT appeared to understand it was a language model somehow.

“ChatGPT understands that at the URL where we find ChatGPT, a large language model such as itself might be found. It correctly makes the inference that it should therefore reply to these questions like it would itself, as it is itself a large language model assistant too,” Degrave wrote on his blog. 

We’re not suggesting ChatGPT has any self-awareness or anything like that. It doesn’t. It’s still mostly a dumb machine that’s good at tricking us it’s smart sometimes.

Cohere launches multi-lingual langugage model

Canadian AI startup, Cohere, has released a model capable of supporting over a hundred different languages.

Most commercial large language model APIs only operate in English, despite advances in machine translation, making it more difficult for developers to build natural language processing-based products for different commmunities.

Cohere has trained its new model on a dataset consisting of over 1.4 billion pairs of text translations of tens of thousands of websites across hundreds of languages.

“This will enable new markets, countries, and global companies to better serve their customers across the globe,” the company declared in a blog post.

Not only will this give non-English speaking companies around the world the ability to perform things like search or sentiment analysis, but also the ability to use different languages for automated content moderation, according to Cohere. ®