A data center rack utilizing living neurons is now functioning in Singapore.
NUS Medicine, data center operator DayOne, and Melbourne's Cortical Labs have introduced a prototype of a biological data center in Singapore, consisting of a 20-unit rack of computers powered by living neurons. The claim regarding its efficiency lacks any published data.
The system, which operates on human neurons, was activated in Singapore and presented to approximately 80 attendees on August 6. The hardware includes 20 CL1 units, which the collaborators claim is the first independently operated biologically integrated server rack globally. The neurons utilized are cultivated from stem cells at the NUS Life Sciences Institute.
The rationale presented for biological computing hinges on electricity consumption. This method is proposed as a means to enhance AI capability while consuming significantly less power than conventional silicon servers. However, the announcement did not provide any figures to substantiate this assertion. According to Cortical Labs' specifications, each CL1 unit consumes between 850 to 1,000 watts, suggesting that the entire rack may draw up to 20 kilowatts.
This power consumption is typical, as a standard data center rack operates in a similar range. The consumption mainly goes toward sustaining the living cells rather than computational tasks, which presents a different challenge from the ones other startups are addressing.
The efficiency lies with the neurons, rather than the infrastructure surrounding them. Switzerland’s FinalSpark, which operates a remote platform featuring 16 brain organoids, claims that its biological processors consume about a million times less energy than traditional digital chips. These assertions center on the cells performing computations. The infrastructure components like incubators, pumps, and temperature controls are the ones that contribute to the electricity costs.
In Europe, a similar goal is being pursued using silicon technology. The Netherlands is developing a neuromorphic computing hub centered on chips that replicate neuron functions without requiring sustenance.
Cortical Labs aims for commercial clients rather than solely focusing on academic research. Founder Hon Weng Chong states that this prototype transitions the dialogue toward commercial use, targeting applications in drug discovery, humanoid robotics, cybersecurity, and fraud detection.
One limitation of this approach is biological; the CL1’s life support system allows its neurons to remain viable for up to six months, necessitating maintenance that silicon systems do not require.
The key metric for comparison would be watts per useful task, which has not yet been published. Meanwhile, investors are continuing to fund brain-inspired alternatives that solely rely on electrical power.
Other articles
Alibaba's profit dropped by 75% as its quarterly expenditures on AI reached $10 billion.
Alibaba's expenditure on AI reached $9.98 billion in the June quarter, while profits declined by 75%. However, the cloud division, for which it incurs costs, achieved a segment profit increase of 133%.
Tracking data indicates that Tesla's robotaxis in Austin are now completely autonomous.
Throughout a two-week period, all 170 Tesla robotaxi trips monitored in Austin operated without supervision. Just two days prior, a passenger captured footage of one navigating through bollards.
Android 17 prefers to throttle one problematic app instead of affecting the performance of your whole phone.
Android 17 implements stricter memory limits for individual apps, which can slow down or stop apps that consume excessive resources. This gives Google a new method to safeguard the overall performance of the phone, preventing a single memory-intensive app from causing broader issues.
The Google Pixel 11 will allow you to tailor the vibration settings of your phone.
Google's concealed vibration options are finally available. The Pixel 11 series allows you to choose distinct vibrations for ringtones, notifications, and even individual contacts.
Android 17 prefers to slow down a single problematic app instead of affecting the performance of your entire phone.
Android 17 implements tougher memory limits for individual apps, which can either slow down or shut down apps that are using excessive memory. This provides Google with a fresh method to safeguard overall phone performance before a single memory-intensive app leads to broader issues.
Octopus Charge consolidates 210,000 US charging stations into a single app, excluding Tesla, Ionna, and Electrify America.
Octopus Energy has introduced Octopus Charge in North America, featuring 210,000 chargers. In Europe, the service boasts 1.5 million chargers, which is attributed to regulatory requirements.
A data center rack utilizing living neurons is now functioning in Singapore.
In Singapore, NUS, DayOne, and Cortical Labs have established a 20-unit rack of biological computers. However, the assertion of its efficiency lacks specific numerical support.