Are we ready for human brain scale AI? – EEJournal

0

My poor old noggin is full of ideas right now that ricochet like corn kernels in an over-enthusiastic popcorn machine. As a tantalizing teaser, did you know that, “Tachyum enables human brain scale AI and moves the whole world into a greener era by delivering the world’s first universal processor”? Otherwise, I’ll be happy to expose, explain and elucidate, but first …

I cringe and currently tear my clothes off because I just missed the opportunity to commemorate my millionth birthday. I celebrated my hundredth birthday last weekend (see also Uh! It’s my 100th birthday!), assuming – of course – that we’re working in base 8 (octal) and not desperately clinging to our grandparents’ base 10 (decimal) number system. But then my friend Bob zeidman pointed out that it was also my 1,000,000th birthday if we choose to work in base-2 (binary).

Well this ship has sailed this year, but now I’m having a palindromic celebration next year when I have 101 (base-8) or 100,0001 (base-2). Of course, 64 is the new 40 if we happen to work in base-16 (hexadecimal), but I’ll keep the hexadecimal in reserve until I’m 66 (base-10), which will be 42 (base – 16). At that point, I have every hope of knowing the answer to life, the universe, and everything.

As an aside, I recently reread Johnny and the dead and Johnny and the bomb by the late, great, Terry Pratchett. Both of these books center on a 12 to 13-year-old boy named Johnny Maxwell who has unusual gifts, like the ability to see the dead (in a good way).

Until last weekend I had no idea these had been made into TV mini-series, the first in 1995 and the second in 2006. I watched Johnny and the dead on Youtube last night while writing these words. On the one hand, it is clear that television techniques and technology have come a long way since 1995; also, it is easy to see that this was a low budget production. On the other hand, the plot of the video was very close to the book and it was great to see the characters I had read come to life, which is ironic in the case of many actors who are somewhat contested in this department.

Now i can’t wait to watch Johnny and the bomb on Youtube, not least because it was shot in 2006 and – based on the clip I just watched – I’m hopeful that the budget and the cinematography will be a treat for my orbs. Additionally, this tale involves time travel (which I love to contemplate) as Johnny and his friends return to World War II to avoid deaths from a misdirected bomb during an air raid.

The reason Johnny Maxwell stories came to my mind is that a recurring character is an old sack lady called Mrs. Tachyon who pushes a supermarket cart while muttering unexpected words and cryptic mumbles that no one else is. understands. Turns out the cart is a time machine, which is a big part of the jars of pickles and fish and chips wrapped in a decades-old newspaper, but “that’s all i have to say about itAs Forrest Gump might say.

What do you mean by “What does all this have to do with anything?” Are you not paying attention? How could a company called Tachyum do not do you think of Mrs. Tachyon?

The reason this all happened is when people from Tachyum contacted me to tell me about the upcoming availability of their FPGA Prodigy Universal Processor FPGA-based emulation prototype. Of course, you might not find this news extremely exciting unless you are already aware of the power of Universal Prodigy processor.

Let’s step back and think about a few trivia, like the fact that data centers currently consume around 3% of the world’s electricity supply (that’s 60% more energy than the UK as a whole). At the current growth rate of 27%, unless something disruptive changes things (like the Prodigy Universal Processor, for example), it will grow to 33% by 2030 and 50% by 2040. Eeek !

Where is all this power going? Well, I don’t know about you, but as a simple example, when my wife (Gina the Gorgeous) and I watch a program on TV, we are constantly checking the facts. Whenever I have a Google Quick, I think of the large number of servers in data centers around the world that are engaged in retrieving the information that I and countless others constantly request. I also have to admit that I feel a little guilty about the amount of energy used to satisfy my requests for non-essential information.

A very good book to help you understand all that is involved in this sort of thing is Tubes: a journey to the center of the Internet by Andrew Blum. And if you really want to get a feel for what’s going on in cyberspace (where no one can hear you screaming), you can check out the Live Internet Statistics website where you can watch a representation of the amount of “stuff” happening on the internet every second (it’s alarming to see the data-seconds go by).

Now, I don’t intend to talk about the Prodigy Universal Processor in depth here because (a) it would require a full-fledged column, and (b) it doesn’t exist yet (more on that momentarily). Suffice it to say, it will be a multi-core device with 16, 32, 64, and 128-core versions in the pipeline (no pun intended). Additionally, each core is claimed to be smaller than an ARM and faster than the fastest Xeon while consuming a tenth of the power.

Targeting large-scale data centers, the architecture of the Prodigy Universal Processor is expected to outperform central processing units (CPU), graphics processing units (GPUs) and tensor processing units (TPUs) for data centers, artificial intelligence (AI) and performance computing (HPC) applications. For example, Prodigy will outperform NVIDIA’s fastest GPU in HPC, as well as AI training and inference tasks (125 Prodigy HPC racks can provide 32 EXAFLOPS tensors). As the people of Tachyum say:

Tachyum’s Prodigy can run HPC applications, convolutional AI, explainable AI, general AI, bio AI, and spiked neural networks, as well as normal data center workloads, on a single platform. homogeneous processor shape using existing and standard programming models. Without Prodigy, data center customers must use a combination of CPU, GPU, TPU, and other accelerators for these different workloads, creating inefficiency, expense, and the complexity of maintaining separate hardware infrastructures. Using specific hardware dedicated to each type of workload (eg, data center, AI, HPC) results in significant underutilization of hardware resources and more challenging programming, support, and maintenance environments. Prodigy’s ability to seamlessly switch between these diverse workloads dramatically changes the competitive landscape and dramatically improves data center economics.

One thing I found particularly interesting is the fact that, as part of its claim to be “the world’s first universal processor,” Prodigy will run the older x86, ARM, and RISC-V binaries in addition to its Prodigy code. native.

Some interesting reads from 2020 tell how Tachyum’s reference design to be used in 2021 AI / HPC supercomputer and Tachyum Joins I4DI to Design World’s Fastest AI Supercomputer in Slovakia. Additionally, just two days ago at the time of this writing, Tachyum announced that his Prodigy Universal Processor successfully upgraded to 5-nanometer process.

Now this is where we have to be a little careful because – if you are visiting the Products on the Tachyum website – your gut feeling might just be that the Prodigy devices are sitting on the shelf waiting for you to order them. In reality, however, we might be better off thinking of it as “a forward-looking leaflet presented for our reading by means of a medium provided on the Web” (sometimes I even wonder myself). What Tachyum currently has is a simulation verified architecture which has now been physically verified using test fixtures.

The last step before full production is physical emulation, which brings us back to the fact that the people of Tachyum are currently bouncing off the walls in excitement about their FPGA’s upcoming availability Prodigy Universal Processor FPGA-based emulation prototype.

Chi To, Director of Solutions Engineering, Tachyum, with the Prodigy Universal Processor FPGA-based emulation prototype (Image source: Tachyum)

The card in the image above measures 14.5 inches by 16 inches (368.3 mm x 406.4 mm). This 24-layer body beauty contains 5,948 components mounted on both sides of the substrate. In particular, the big four components with cooling fans are Intel Stratix 10 GX FPGAs, each containing more than 10 million logical elements (LE). This single board can be used to emulate eight Prodigy processor cores, including their fixed and floating point vector and matrix processing units. The complete hardware emulator consists of multiple FPGA and I / O boards connected by cables in a rack.

In addition to allowing Tachyum chaps and chapesses to perform a final check before full production release, customers will be able to use Prodigy’s fully functional FPGA emulation for product evaluation and performance measurements, as well as for software development, debugging and compatibility testing. The Prodigy FPGA emulation system will help customers smooth the adoption curve of Prodigy in their existing or new data center and / or HPC systems that demand the combination of high performance, high utilization and reliability. low power consumption.

The human brain is an awesome organ (at least mine is) that contains between 100 and 150 trillion neural connections (synapses) – say 125 trillion “give or take” to provide a point of reference. Just counting the connections is a meaningless exercise, but the meaningless exercises are what I do best. In 2018, Google introduced an AI model called BERT with 340 million connections for use in natural language processing (NLP). Now, just three years later, models like Open AI’s GPT-3 have up to 175 billion connections, which is enough to act as a smart chatbot. By 2023, it is predicted that high-end AI models will have 100,000 billion connections capable of performing more than 100 AI exaflops.

What we’re talking about here is human brain scale AI. The more I think about it all, (a) the more I’m afraid (see also The Artificial Intelligence Apocalypse: Is There Time To Be Afraid?) and (b) plus I think the guys and girls at Tachyum are in the right place at the right time with their Prodigy Universal Processor. What you say? Are you excited or terrified of what seems to be heading our way?


Source link

Share.

About Author

Leave A Reply