Tesla Begins to Produce Dojo Supercomputer to Practice Driverless Automobiles

In its second quarter earnings report for 2023, Tesla has introduced the graduation of manufacturing for its Dojo supercomputer, a important part for coaching its fleet of autonomous automobiles. The corporate outlined 4 essential expertise pillars mandatory to realize automobile autonomy at scale: an intensive real-world dataset, neural web coaching, automobile {hardware}, and automobile software program, and emphasised that they’re internally growing every of those pillars.

‍

The Dojo coaching laptop represents a major development in direction of sooner and less expensive neural web coaching. Whereas Tesla already possesses a robust Nvidia GPU-based supercomputer, the Dojo custom-built laptop makes use of chips particularly designed by Tesla. CEO Elon Musk had beforehand dubbed this high-performance coaching laptop as “Dojo,” envisioning it to be able to an exaflop, equal to 1 quintillion (10^18) floating-point operations per second. This immense computing energy is troublesome to fathom, as it could take over 31 billion years to carry out the identical variety of calculations sequentially.

‍

Tesla’s progress on the Dojo mission was shared throughout its AI Day in 2021, the place executives unveiled its preliminary chip and coaching tiles, which had been slated to type an entire Dojo cluster or “exapod.” The plan concerned combining two units of three tiles in a tray and putting two trays in a pc cupboard, leading to over 100 petaflops per cupboard. With a 10-cabinet system, Tesla aimed to interrupt the exaflop compute barrier with its Dojo exapod.

‍

Subsequent updates at AI Day 2022 showcased additional developments, together with the presentation of a full system tray for the Dojo. Tesla had beforehand projected a full cluster by early 2023, although the newest reviews point out that the completion of the Dojo exapod will doubtless happen in early 2024.

‍

Tesla’s dedication to growing its Dojo supercomputer demonstrates its dedication to pushing the boundaries of autonomous automobile expertise by progressive in-house options. By leveraging this highly effective software, the corporate goals to make important strides in coaching its fleet of autonomous automobiles, in the end driving developments within the area of car autonomy at scale.

‍