Sesha 🦇🔊

AI, Blockchain and Sustainable Abundance

[Sesha 🦇🔊 @absoluteSesha]

@teslayoda Google will just match the price. It wont even make a dent in google earnings. For Tesla on the other hand the PE is significantly high so keeping Robotaxi rides cheap will have a direct affect on stock price

[Sesha 🦇🔊 @absoluteSesha]

@davelee V15 + AI4.5 is unsupervised FSD which will be good enough to drive in any condition. From comments in latest earnings call it looks like V15 will be fully done by end of this year or early next. Only minor improvements from there are needed.

[Sesha 🦇🔊 @absoluteSesha]

@DillonLoomis Thank you for the honesty. You have more of trust now than before.

[Sesha 🦇🔊 @absoluteSesha]

@Amy787 My take: CyberCab and other Tesla models can use AI4.5 for unsupervised driving. Chips are a constraint for Optimus full functionality but for the first few years where the roll out could be industrial and special purpose than general, AI5/6 would do. Terafab is needed for 2030+

[Sesha 🦇🔊 @absoluteSesha]

@nymbusjp waymos do respond to temporary traffic diversions not perfectly yet though. my overall point is : all autonomous systems are work in progress currently and there is a path for google to execute on without hd maps even if they keep using lidar.

[Sesha 🦇🔊 @absoluteSesha]

@nymbusjp Great article!! But I think it cannot be assumed that advanced AI models can’t solve sensor fusion problem. Camera vision system would be central to any system though. But my original point was we cannot assume HD maps is central to Waymo going forward or even currently.

[Sesha 🦇🔊 @absoluteSesha]

@stevenmarkryan Giving up on model development?

[Sesha 🦇🔊 @absoluteSesha]

@nymbusjp Not necessarily. Data from different sensors can be provided as input and an AI model can output control data as output. Getting depth information from sensors could lighten the load for AI processing. Would also eliminate grey kitten on grey tarmac problem.

[Sesha 🦇🔊 @absoluteSesha]

@MatthewBerman @bot Grok bot for this human bot please..

[Sesha 🦇🔊 @absoluteSesha]

@nerdalert or AI4.5 or AI5 is a must to make any more progress towards autonomy and it needs new data and new training runs to take advantage of new hardware.

[Sesha 🦇🔊 @absoluteSesha]

@beffjezos they took forever to build 4680s. on that timeline this is a 2040 thing.

[Sesha 🦇🔊 @absoluteSesha]

@elonmusk @shivon what about first principles when dressing?

[Sesha 🦇🔊 @absoluteSesha]

@DennisCW_ model 3 is about to be killed. not need for it in robotaxi future

[Sesha 🦇🔊 @absoluteSesha]

@teslayoda magic of “partnership”. tesla will spend the money, spacex will enjoy the benefits.

[Sesha 🦇🔊 @absoluteSesha]

Better to have subdued expectations than be disappointed. AI5 is a priority for a reason. Cybercab might not go wide without AI5. Units currently being built might be collecting training data using new cameras and bigger model(s) will be deployed on AI5 before going wide.

[Pejjy @CuriousPejjy]

Now that $TSLA has officially launched the Unsupervised Cybercab Robotaxis for employees only, I expect a public launch as soon as 3 weeks and no later than September.

Tesla has tested the Cybercab for months now and launching it for employees indicates that they are extremely confident to launch it to the public soon. Every time when Tesla wants to push a major software product out, they always launch it for employees first meaning the product is ready and just needs to be tested one last time!

The 2nd half of 2026 is going to WILD.

[Sesha 🦇🔊 @absoluteSesha]

@JoeTegtmeyer steering wheel?

[Sesha 🦇🔊 @absoluteSesha]

@RangarajSrikan1 True. Tesla is not competing with legacy players anymore. Considering AI5 and dojo hiccups it is way better to acquire and move fast.

[Perplexity @perplexity_ai]

Introducing Perplexity Computer.

Computer unifies every current AI capability into one system.

It can research, design, code, deploy, and manage any project end-to-end.

[Sesha 🦇🔊 @absoluteSesha]

@eivissacopter so this new pack would not be structural?