Tesla Talk 2

That's quite literally what the most expensive dashcams do hence they carry price tags of 500 euro plus. They work with RAW or LOG footage or offer lower compression. Most folks aren't going to pay that so cheaper stuff compresses the imagery for higher storage capabilities. Again most consumers prefer more record time or a balance mix between the two.

But as I said he's talking absolute bluster. This is doing nothing that Sony hasn't been doing with their sensors for years. And most new high end smartphones can be switched over to store in these formats for processing footage later on.

This isn't offering anything new or groundbreaking but he is trying to make people think it is. For what purpose I've no idea but he's pretty boring at this point with his limited grasp of the subjects and bluster in which he describes mundane technical tasks that are extremely common. This isn't Tesla specific, it isn't even specific to this use case either.

"Look we count the photons and amp up the giga spectrum to enhance the gamma spectrum so we have infrared vision capabilities" (or something to that effect) . Yes I understand this is boring stuff to most but it's the reason I'm annoyed by the marketing of said features.

The actual interesting stuff is in the hardware and what it's doing with the imagery from a decision capability and what is onboard versus what's being off boarded. (Cloud compute)

Other manufacturers most definitely are storing the files in LOG or RAW and processing off of that , they are not 'following suit' this is literally the standard in the industry of image processing.

Even on the basic smell test this stuff falls short and it's dishonest.

Tesla's prowess is Not in this. Their prowess is on the LLM use in image processing and the volume they've processed at. Nothing whatsoever to do with the methodology of storing and working with raw data sets. That's simply not true.

Can we at least deal in facts here. They do a lot of good things without this stuff being fluffed up as ground breaking.

Bent on Tesla destruction? No I've never once ever inferred that. In fact quite the opposite. I have said what I think they need to be doing to compete and excel against their competitors. Something they are more than capable off. Whilst jettisoning the bobble head.

Fyi 'neural net' as fluffy calls it is actually LLM. He bastardised the technology terms to make them sound fancy to investors as LLM let's face it is boring. It's similar type of architectural make-up that is used by Google for their image analysis. There is nothing neural about it. It's simply LLMs working together on task specific roles.

Id suggest avoid getting grok to summarise Elons press releases.

Tesla have a fantastic analysis engine in fact it perhaps is one of their greatest in house IP. But it can only get better within a combined system.

We're getting into symantics here your describing the layman term of how LLMs engage with each other architecturally. This is basic LLM that is used across the AI industry. No one is calling this by it's marketing term. Common parlance would actually be AI as it's the buzzword. Elons sticking with Neural because it's helpful for him to pretend it's differentiator. (Marketing)

I digress, we are widely down a path again for no gain.

I see Rivian are in talks with building in house Lidar to circumvent tariff and other restrictions. Also vertical integration prospects. A smart move on the face of it .

He's not and I do. He doesnt describe things in technical detail he uses superlatives consistently. anyway lets move on

__

Michael Martin seems to think so.

It appears on the face of it they're not. Musk signed a deal last week with his arch nemesis on anthropic to lease out their infrastructure to anthropic. He had alot of unused space which he wasn't scaling Grok into, there seems to be somewhat of a switch away from xAI being a core product. It's unclear why though. But the anthropic deal is a big one, that's capacity that was ear marked for xAI originally.

Haha what!? Tell me more!

I found out another Science channel I follow, Veritasium, sold out to some private equity crowd a while ago too.