https://www.xataka.com/aplicaciones/desafio-imposible-neuralink-no-implante-cerebral-uno-que-nos-recuerda-mucho-a-pied-piper-a-miticos-mp3

The startup created by Elon Musk has made dramatic progress in recent months: The first human implant appears to have been successful despite some problems, but it faces a huge challenge. It’s something we can’t imagine, and it has a lot to do with MP3 files.

Mission impossible? Judging by the requirements, it certainly looks like it. They announced a “compression challenge” on Neuralink, with the goal of reaching a 200:1 compression ratio. Use only one bit in the compressed file for every 200 original bits.

And lossless. Not only that: the compression must be carried out in real time (latency less than one ms) and also with minimal energy consumption (less than 10 mW including data transmission via radio). The compression algorithm should result in a lossless compressed file; This means that uncompressing the file will recover all data in the original. There is more information about the issue in the GitHub repository.

Too much data per second. According to this problem definition, the Neuralink N1 implant produces approximately 200 Mbps of electrode data (1,024 electrodes at 20 kHz with 10-bit resolution) and can transmit wirelessly at approximately 1 Mbps. This means that more than 200 times compression is required to solve the problem posed by Neuralink.

Reference data. To work on the issue, Neuralink is making a 143 Mbyte ZIP file available to interested parties. This corresponds to recording natural signals from the electrodes of a Neuralink implant implanted in a primate. The recording was made while the animal was playing a video game; This is something we have seen in the past.

Noise (visible) everywhere. According to some users’ comments on Twitter, the initial data displayed in a spectrogram shows a very large amount of noise, at least in appearance. It is possible to compress data if it contains (almost) no noise, for example if patterns are found in the data. That’s what the Huffman algorithm does, but it seems even harder to do with all this noise.

A problem worth billions. Other users point out something even more important: This level of compression without loss and energy consumption will be extraordinary not only for Neuralink, but also for countless applications. For example, for large online data storage centers, they will suddenly see how they can compress all this information to a 200:1 ratio.

But Neuralink doesn’t even mention the economical offer. Interestingly, there are no details about a potential financial reward in the challenge post. They only state that anyone who finds the solution should contact Neuralink via their email address.

Hello Pied Piper. The famous TV series ‘Silicon Valley’ in its first season showed how the lead character Richard Hendricks invented an application called ‘Pied Piper’, which uses an amazing compression algorithm. HBO producers even asked a Stanford professor to create an actual metric (Weissman score) that would show how good each compression algorithm is.

compression fever. In reality, the problem posed by the great series is very real, and compression algorithms have been trying to go further for decades. The funny thing is that since ‘Silicon Valley’ was released, every new achievement in this field (there are frequent hackathons on this task) is celebrated as if it were a new iteration of the Pied Piper.

Long live MP3s. In the late 1980s, mathematician Karlheinz Brandenburg complemented the work of Eberhard Zwicker and Dieter Seitzer, who were working on high-quality music compression. Both realized that CDs stored data that even humans could not capture, and together with Brandenburg, they developed a format they called MP3. This technology eliminated this unnecessary data, allowing songs to be compressed at a 12:1 ratio. For each original 12 bit, the MP3 format only needed one. The big difference? This MP3 is a lossy format and this will not work for Neuralink.

Image | HBO

in Xataka | Neuralink will be eight years old, and its biggest challenge so far hasn’t been innovation. catching up