Human brain organoid bioprocessors now available to rent for $500 per month
FinalSpark estimates its Neuroplatform is a million times more energy efficient than a digital processor.
FinalSpark, the firm behind Neuroplatform, has begun to offer paid 24/7 remote access to its bioprocessors. In May, we reported on these pioneering human brain organoid-based processors and their touted million times greater power efficiency when compared to digital processors. Now we note that academic customers can get access to this biocomputing platform, featuring four shared organoids, for $500 per user per month (or even free, for selected projects).
For the fee, FinalSpark says that users get to conduct biocomputing research on a 24/7 fully managed remote neuroplatform. An infographic shows a remote researcher interfacing with a ‘Forebrain Organoid.’ Furthermore, users are promised:
- Integrated R&D environment for biocomputing research
- Realtime neural stimulation and reading
- Programming API for Python
- Digital notebook for documentation and research
- Data storage and backup
- Technical support
Neuroplatform claims to be the world’s first online platform delivering access to biological neurons in vitro. The neuron-packed organoids are capable of learning and processing exceptionally efficiently. FinalSpark hopes that adopting bioprocessors based on biological neurons rather than transistors could significantly impact the incredible energy expenditure we often hear about in the tech world. Saving billions of watts when training LLMs or other intensive tasks should also be a positive for the environment.
It is worth recapping how the Neuroplatform works. The wetware architecture mixes hardware, software, and biology using a quartet of Multi-Electrode Arrays (MEAs) housing human brain organoids in a microfluidic life support system. 3D tissue masses are interfaced and stimulated by eight electrodes, with monitoring cameras and a tuned software stack so that researchers can input data variables and read and interpret processor output.
The information available about FinalSpark’s Neuroplatform has grown in recent months on the journey to remote rental availability. A blog post in July provided a closer look at how the labs create organoids and how the researchers are sure they are packed with neurons. An earlier blog post shared a macrograph of a single brain organoid, reproduced above, which is estimated to contain 10,000 living human neurons.
Five major institutions are collaborating on the Neuroplatform, with nine listed users. The platform has been opened to broader academia, which can rent access to the biocomputer for $500. It will be interesting to see what results from the platform’s use and how it will develop.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Mark Tyson is a news editor at Tom's Hardware. He enjoys covering the full breadth of PC tech; from business and semiconductor design to products approaching the edge of reason.
-
Elf_Boy I question how using human brain tissue is ethical?Reply
How long does it survive?
If they keep making bigger organoids when do they start to think?
No doubt there are answers to some/all of these questions. I am interested in seeing them. -
bit_user
I think these are for research and not a substitute for digital artificial neural networks. AFAIK, there's no way to read out the weights from one organoid or program them into another. That's a key capability computers offer that biology simply lacks. It takes like 10k GPU hours to train a huge LLM, but then you can replicate it and deploy it for inferencing very cheaply.The article said:FinalSpark hopes that adopting bioprocessors based on biological neurons rather than transistors could significantly impact the incredible energy expenditure we often hear about in the tech world. Saving billions of watts when training LLMs or other intensive tasks should also be a positive for the environment.
Also, you surely know how hard it is to learn even a fraction of what some of these popular LLMs know. And that's using your whole human brain. There's no way an organoid is going to learn data on the scale of what LLMs deal with.
Finally... oh I must've forgotten!: D -
bit_user
Or maybe your training drives it insane?usertests said:What if it goes on strike? :P
...but seriously, I'm sure it needs to sleep, right? Sleep is a biological requirement of real neurons. I'm pretty sure everything with a brain or central nervous system sleeps - even insects. -
usertests
I think we'll end up seeing brain-computer hybrid technology to get the best of both worlds, IF literal biology is going to play any significant role at all in future computing/AI. Organoids could also be run in parallel with fast interconnects.bit_user said:Also, you surely know how hard it is to learn even a fraction of what some of these popular LLMs know. And that's using your whole human brain. There's no way an organoid is going to learn data on the scale of what LLMs deal with.
Brain-inspired neuromorphic computing can also be a lot more efficient than AI-focused GPUs. But it won't die if you don't give it nutrients on schedule.
We haven't even scratched the surface of what's possible, because unlike organoids, we don't have massively multi-layered monolithic 3D chips, neuromorphic or otherwise (if the neuron-like elements use infrequent spikes like neurons, that lowers power consumption and heat, which is one of the biggest issues with 3D).
https://en.wikipedia.org/wiki/Cerebral_organoidElf_Boy said:I question how using human brain tissue is ethical?
Brain organoids are permitted in part because they are small and simple, avoiding the ethics issues. You have to wonder if connecting them in parallel or some other tricks could increase their capabilities.
I've heard 1 year for survival but they may have extended that. -
bit_user
You're ignoring the elephant in the room, which is how to read out & program the weights. Unless/until you solve that, I see these organoids being useful only for a limited range of research projects.usertests said:I think we'll end up seeing brain-computer hybrid technology to get the best of both worlds, IF literal biology is going to play any significant role at all in future computing/AI. Organoids could also be run in parallel with fast interconnects.
Brain-inspired neuromorphic computing can also be a lot more efficient than AI-focused GPUs. But it won't die if you don't give it nutrients on schedule.
I have no idea why you think the physical structure of chips needs to match that of a 3D brain. Computer memory is effectively 1-dimensional (it all gets mapped into a linear address range), but we can use it to represent & process N-dimensional data.usertests said:We haven't even scratched the surface of what's possible, because unlike organoids, we don't have massively multi-layered monolithic 3D chips, neuromorphic or otherwise (if the neuron-like elements use infrequent spikes like neurons, that lowers power consumption and heat, which is one of the biggest issues with 3D).
Sure, if you have a distributed memory or processing-in-memory architecture, then you can derive better power-efficiency by employing chip-stacking, but it's really just a power-saving technique. Any network topology can be processed as a graph, which you can map to a linear or planar array of processors. -
ThomasKinsley
Either this is an elaborate hoax or this is an ethical scandal of epic proportions. The fact that this has played out many times in movies as a trope and nobody is raising ethical concerns in these articles leads me to believe the former is true, but get me off this world if it's the latter.Elf_Boy said:I question how using human brain tissue is ethical?
How long does it survive?
If they keep making bigger organoids when do they start to think?
No doubt there are answers to some/all of these questions. I am interested in seeing them.