https://fabrizio-ottati.github.io/tags/research/Recent content in Research on Fabrizio OttatiHugoen-USTue, 26 Dec 2023 00:00:00 +0000https://fabrizio-ottati.github.io/blog/northpole/Tue, 26 Dec 2023 00:00:00 +0000https://fabrizio-ottati.github.io/blog/northpole/<figure><img src="https://fabrizio-ottati.github.io/blog/northpole/on-chip-cartoon.png" width="600px"> </figure> <p><a href="https://research.ibm.com/blog/northpole-ibm-ai-chip">NorthPole</a> is the new shiny artificial intelligence (AI) accelerator developed by IBM.</p> <blockquote> <p>NorthPole, an architecture and a programming model for neural inference, reimagines (Fig. 1) the interaction between compute and memory by embodying 10 interrelated, synergistic axioms that build on brain-inspired computing.</p> </blockquote> <p>The creators claim that NorthPole design has been driven by 10 axioms inspired by human biology. We will use them as guidelines to analyze the paper, keeping the same outline as the original article.</p>https://fabrizio-ottati.github.io/draft/northpole/Sat, 25 Nov 2023 00:00:00 +0000https://fabrizio-ottati.github.io/draft/northpole/<p><a href="https://research.ibm.com/blog/northpole-ibm-ai-chip">NorthPole</a> is the new shiny artificial intelligence (AI) accelerator developed by IBM. The creators claim that NorthPole design has been driven by 10 axioms inspired by human biology. We will use them as guidelines to analyze the paper.</p> <p>The outline of this blog post is the same as the original article.</p> <h1 id="axiomatic-design">Axiomatic design</h1> <blockquote> <p>NorthPole, an architecture and a programming model for neural inference, reimagines (Fig. 1) the interaction between compute and memory by embodying 10 interrelated, synergistic axioms that build on brain-inspired computing.</p> </blockquote>https://fabrizio-ottati.github.io/blog/truenorth/Sun, 26 Mar 2023 00:00:00 +0000https://fabrizio-ottati.github.io/blog/truenorth/<p><img src="brain-to-chip.png" alt="The beautiful TrueNorth article image"></p> <h2 id="why-do-we-want-to-emulate-the-brain">Why do we want to emulate the brain?</h2> <p>If you have ever read an article on neuromorphic computing, you might have noticed that in the introduction of each of these there is the same statement: &ldquo;The brain is much powerful than any AI machine when it comes to cognitive tasks but it runs on a <strong>10W</strong> power budget!&rdquo;. This is absolutely true: neurons in the brain communicate among each other by means of <strong>spikes</strong>, which are short voltage pulses that propagate from one neuron to the other. The average spiking activity is estimated to be around <strong>10Hz</strong> (i.e. a spike every 100ms). This yields <strong>very low processing power consumption</strong>, since the activity in the brain results to be <strong>really sparse</strong> (at least, this is the hypothesis).</p>