A rough estimate shows that human brain is 30 times faster than best of supercomputers
Scientists have developed a new method to compare supercomputers to human brains wherein the computer’s performance was compared to a rough estimate of how frequently the neurons of the brain is able to fire off electrical signals.
In July this year, Silicon Valley pioneer, Elon Musk, founder of Tesla Motors, SpaceX and his Open Philanthropy Project announced a funding of $7 million among the 37 research teams which would be working on the ‘Killer AI issue’. These 37 research teams will work on the AI issue from different angles which would include teaching AI to make out what humans want from them, also aligning robot’s interest with that of humans and keeping a complete human control over the AI.
Basically, Musk aims to guide the development of smarter AI with least potential threat to humanity.
One of the 37 research projects, funded by Musk is known as “AI Impacts project” and it aims at finding a newer technique which can compare supercomputers to the human brain.
Instead of focusing on how fast the hardware or brain can do certain calculations, the research team here developed a method through which they determined which among the two, brain and computer, is able to communicate the messages instantly within its own network. Based on this experiment the researchers plan to set a benchmark which will provide some useful way to compare the AI’s progress with the human intelligence.
AI Impacts Project:
AI Impacts project emerged from the brains of two PhD students, one from the University of California, Berkeley and another from Carnegie Mellon University. Basically, these scientists developed a new technique which can be used to compare human brains to supercomputers.
The preliminary methodology termed as Traversed Edges Per Second (TEPS) measures the speed at which a computer can transmit information within its own system. A computer needs to simulate a graph and then search through it, based on this the typical TEPS benchmark has been set. However, a brain cannot simulate a graph like computer. Hence, researchers had to compare the performance of computer with the rough estimation of how frequently the neurons of the brain is able to fire off electrical signals.
“A big pragmatic benefit of measuring the brain in terms of communication is that it hadn’t been done before,” says Katja Grace, a researcher at the Machine Intelligence Research Institute in Berkeley who is working on a doctorate in Logic, Computation, and Methodology at Carnegie Mellon University. Grace added that this method “provides a relatively independent estimate of the price of computing hardware roughly comparable to the brain.”
Boston-based Future of Life Institute has granted a funding of $49,310 to these two researchers for their AI Impacts project and thus have been listed under the group of Elon Musk’s AI research project.
Jeremy Hsu at IEEE Spectrum says:
“IBM’s Sequoia supercomputer currently holds the TEPS benchmark record with 2.3 x 1013 TEPS. Grace and her collaborator, Paul Christiano, a PhD student in theoretical computer science at Cal Berkeley, calculated that the human brain should be at least at least as powerful as Sequoia at the lower end of their TEPS estimates. At the upper end, their max estimate of the human brain’s capabilities suggest that it’s 30 times as powerful as IBM’s number cruncher at 6.4 times 1014 TEPS. They’ve pegged the cost of the human brain’s performance at somewhere between $4,700 and $170,000 per hour in terms of current computer prices for TEPS. Grace and Christiano say they previously came up with a “fairly wild guess” that TEPS prices could improve by a factor of 10 every four years. That means computer hardware costing $100 per hour to operate could become competitive with the human brain during a time period between seven to 14 years.”
However, there is no need to worry yet, because there is no much information as to how quickly the computer hardware will be able to evolve in its TEPS performance. Researchers also point out that it could be a case that instead of faster evolution the computers might slow down in their TEPS performance. Thus researchers ensure the masses that it is not possible for AI to replace humans any time soon since all these assumptions have a lot of uncertainties packed in their calculations.
Researchers further also add that in case say the ‘TEPS price goal’ proves to be 100% accurate even then one need not worry, because along with a competent computer hardware one would also need appropriate software which will enable the AI to emerge as a more powerful AI and actually compete with human intelligence. Grace pointed out “A laptop’s worth of computing power doesn’t automatically spawn Microsoft Word.”
“We have very little idea how efficiently the brain uses its computational resources, and how that will compare to the efficiency of systems that humans design,” Grace said. “So even if we knew how much hardware was needed to do what the brain is doing in the way the brain is doing it, this might be very different from the amount of hardware human engineers need to achieve the same functions once they have any way to achieve those functions.”
Researchers believe it is easier to measure the communication within the brain’s neurons rather than measuring the computations. Anyways, no one even knows how to measure the computations in the brain because it has not been discovered yet and nobody knows how a computations is represented in the brain …may be after some years humans might even find this secret. However, for now TEPS benchmark seems to be a promising method with which researchers can compare AI with human level intelligence in the near future.
Aim of AI Impacts project:
As mentioned above the major aim of AI Impacts project is to set a new methodology which can help comparing AI’s progress with the human brain.
The other aim of pioneers of this project, Grace and Christiano, is to find out if AI research makes “abrupt and surprising progress at any point” or would it just progress gradually by making small and incremental steps.
Researchers hope the latter proves true because this will give them the required time to predict the AI progress.
Grace thus says: “In aid of this, we are looking at other technologies that have seen abrupt progress, which is interesting in itself. The biggest jump in any technological trend that we have found was from nuclear weapons.”
“AI Impacts Research Bounties”
Researchers at the AI Impacts research want people to give several inputs which would help them in their research work. They are also offering rewards which do not have any specific deadlines. This project has been introduced under “research bounties” and the rewards range from $50 to $500 under the category for someone who provides the example of “discontinuous technological progress”, another category offers $20-$100 for some apt example wherein people acted to prevent a risk that was some 15 years away.
As of now, Grace and Christiano are also studying the Bitcoin hardware to make out the stepping stone which can help in speeding the rate at which hardware improves. Both of them also assume that a comprehensive use of AI can also prove to be tremendously useful in research progress.
In all the researchers aim to build “a quantitative model of how fast artificial intelligence research should be expected to grow in an economy with increasing quantities of artificial intelligence available to do research.”