TechArena’s Three Innovations You Need to Know about Heading into 2023
The last few weeks have been rich with new technology announcements, some with a lot of fanfare and others that happened in the corners of the tech arena that will grow in importance next year. Three grabbed my attention as topics that should squarely fit on your radar as tech innovations that will disproportionally reshape technology as we know it. Let’s take a look.
The new Google? OpenAI introduces ChatGPT (version 3.5)
We have all interacted with chat bots, and to interact with chat bots is to sometimes be infuriated by them. The technology simply has not advanced enough to mimic human thought, asking follow up questions, delivering detailed answers, and because of that we often find ourselves caught in a loop of a very predictable data decision tree that doesn’t deliver. I love Alexa, but she still can’t answer what the most important news of the day is, summarize abstract thoughts, or suggest ideas other than to place items in my Amazon shopping cart. I’m seeking a deeper relationship that at this point is very much one way.
Enter ChatGPT from the OpenAI team. This new AI model feels like it may be finally delivering on what we have been seeking all along…and more. Human-like communication that can deliver detailed answers, seek follow-up, and admit mistakes. This will be a step forward for consumer facing applications, but there’s more to the story. ChatGPT has shown promise to code software, write prose in the fashion of a specific author, and orchestrate cloud instances. The breadth of application is inspiring as the community starts integrating this technology into development streams, and the acute developer interest suggests that this will happen quickly. That doesn’t mean that there won’t be bumps in the road. Stack Overflow almost immediately suspended ChatGPT due to challenges with incorrect information in queries overwhelming the site. Others have called out that ChatGPT lacks morals, doesn’t necessarily like humans, and has an equal ability to do things like create phishing scams and Malware as software for good intentions. More alarmist views express that it will re-define our markets and eliminate endless jobs, which certainly could be the case with this and other AI technology long term but likely won’t happen in week one of implementation.
The net takeaway? This is a powerful tool and integration into business could provide incredible advancement making it a technology and a space with expected future innovation not to be ignored.
Meta’s new Protein-Folding AI will Re-Shape Life
OK, maybe that sub-head was a bit hyperbolic, but Meta’s introduction of the ESM Metagenomic Atlas is an important follow on from DeepMind’s 2020 announcement of AlphaFold2 in massively accelerating protein shape predictions at the molecular level. First discussed on the TechArena by VAST Data’s Jeff Denworth, AlphaFold2 has had an incredible impact on the scientific community. Institutions like the Max Plank Institute have already come out stating that these tools have solved protein structure challenges that traditional science was failing to unlock, and Amazon made a splash earlier this year with the introduction of AWS Batch providing the computing power at scale to fuel research.
What’s the impact? Researchers can apply this to understand the human genome at a level that has been out of reach accelerating treatments for disease, new vaccinations and other medical breakthroughs. The Metagenomic Atlas has already mapped over 600 million proteins in its database driving National Geographic to name it as one of the top 22 innovations in 2022. Science magazine had already named AlphaFold2 and other AI driven research in this space as 2021’s breakthrough of the year.
The dystopian thinkers will connect the dots that all of the large cloud players are racing to control the foundations of life and the power that comes with such knowledge. They may also reflect that Meta is working double-time to re-shape its image towards the metaverse and this advancement places them very far away from Facebook likes. For me, these concerns are far outweighed by the fact that both Google and Meta have made these tools open source for the scientific community to integrate into labs immediately, and we all stand to benefit from the accelerated innovation from their collective discoveries. Watch this space.
A Sledgehammer to Moore’s Law Limitations
You’ve read up to this point and thought, Allyson I’ve already heard of chatbots and protein folding. True! These have garnered a boatload of attention in their respective circles. But my third technology to watch heading into 2023 is one that quietly launched at Supercomputing last month, the Universal Chiplet Interconnect Express, or UCIe specification. I first wrote about it in the summary of SC22’s advancements and have since got even more excited about its promise after talking to its architects. So what is UCIe, and why am I holding it in such high esteem?
To answer this question, we need to first look at the problem that UCIe and other technologies like the Compute Express Link (CXL) are trying to solve. Moore’s Law is running out of gas, and shrinking process technology, the foundation for computing performance advancement since the birth of the microprocessor, will only get us so far in advancing semiconductor density. While CXL has provided a fantastic industry standard to connect chips on a computing platform and has eliminated historical proprietary interconnect schemes that have gated true heterogeneous computing, UCIe is going a leap further enabling this same industry standard foundation on chip package. What does this mean? Chiplet architectures, first envisioned by DARPA’s CHIPS program back in 2017, can finally be advanced without the limitation of proprietary technology. This means that both industry and consumers of computing can dial in the right balance of CPU, GPU, DPU etc chiplets into a a microprocessing package to deliver the best performance capabilities for targeted workloads. The who’s who of semis have engaged in the UCIe specifications both from a standpoint of silicon suppliers (enter AMD, ARM, Intel, NVIDIA Samsung, TSMC and more) and large service providers with their own silicon aspirations (hello Alibaba, Google, Meta, and Microsoft). Notably absent is Amazon, and we’ll need to watch to see if they really want to fight an standard consensus that seems destined to take off in products across the industry.
So why am I elevating UCIe to this lofty position? Microprocessor advancement is still foundational to breakthroughs of everything that runs on silicon. Without access to more performance our advances will be gated to today’s processing capabilities, and the computing industry sits alone in terms of the rate of innovation enjoyed for decades based on silicon advancement. Further breakthroughs simply require it, and for me that is a big call for applause for the team who delivered the UCIe 1.0 specification. Expect more details on the TechArena soon. As always, thanks for engaging – Allyson.
The last few weeks have been rich with new technology announcements, some with a lot of fanfare and others that happened in the corners of the tech arena that will grow in importance next year. Three grabbed my attention as topics that should squarely fit on your radar as tech innovations that will disproportionally reshape technology as we know it. Let’s take a look.
The new Google? OpenAI introduces ChatGPT (version 3.5)
We have all interacted with chat bots, and to interact with chat bots is to sometimes be infuriated by them. The technology simply has not advanced enough to mimic human thought, asking follow up questions, delivering detailed answers, and because of that we often find ourselves caught in a loop of a very predictable data decision tree that doesn’t deliver. I love Alexa, but she still can’t answer what the most important news of the day is, summarize abstract thoughts, or suggest ideas other than to place items in my Amazon shopping cart. I’m seeking a deeper relationship that at this point is very much one way.
Enter ChatGPT from the OpenAI team. This new AI model feels like it may be finally delivering on what we have been seeking all along…and more. Human-like communication that can deliver detailed answers, seek follow-up, and admit mistakes. This will be a step forward for consumer facing applications, but there’s more to the story. ChatGPT has shown promise to code software, write prose in the fashion of a specific author, and orchestrate cloud instances. The breadth of application is inspiring as the community starts integrating this technology into development streams, and the acute developer interest suggests that this will happen quickly. That doesn’t mean that there won’t be bumps in the road. Stack Overflow almost immediately suspended ChatGPT due to challenges with incorrect information in queries overwhelming the site. Others have called out that ChatGPT lacks morals, doesn’t necessarily like humans, and has an equal ability to do things like create phishing scams and Malware as software for good intentions. More alarmist views express that it will re-define our markets and eliminate endless jobs, which certainly could be the case with this and other AI technology long term but likely won’t happen in week one of implementation.
The net takeaway? This is a powerful tool and integration into business could provide incredible advancement making it a technology and a space with expected future innovation not to be ignored.
Meta’s new Protein-Folding AI will Re-Shape Life
OK, maybe that sub-head was a bit hyperbolic, but Meta’s introduction of the ESM Metagenomic Atlas is an important follow on from DeepMind’s 2020 announcement of AlphaFold2 in massively accelerating protein shape predictions at the molecular level. First discussed on the TechArena by VAST Data’s Jeff Denworth, AlphaFold2 has had an incredible impact on the scientific community. Institutions like the Max Plank Institute have already come out stating that these tools have solved protein structure challenges that traditional science was failing to unlock, and Amazon made a splash earlier this year with the introduction of AWS Batch providing the computing power at scale to fuel research.
What’s the impact? Researchers can apply this to understand the human genome at a level that has been out of reach accelerating treatments for disease, new vaccinations and other medical breakthroughs. The Metagenomic Atlas has already mapped over 600 million proteins in its database driving National Geographic to name it as one of the top 22 innovations in 2022. Science magazine had already named AlphaFold2 and other AI driven research in this space as 2021’s breakthrough of the year.
The dystopian thinkers will connect the dots that all of the large cloud players are racing to control the foundations of life and the power that comes with such knowledge. They may also reflect that Meta is working double-time to re-shape its image towards the metaverse and this advancement places them very far away from Facebook likes. For me, these concerns are far outweighed by the fact that both Google and Meta have made these tools open source for the scientific community to integrate into labs immediately, and we all stand to benefit from the accelerated innovation from their collective discoveries. Watch this space.
A Sledgehammer to Moore’s Law Limitations
You’ve read up to this point and thought, Allyson I’ve already heard of chatbots and protein folding. True! These have garnered a boatload of attention in their respective circles. But my third technology to watch heading into 2023 is one that quietly launched at Supercomputing last month, the Universal Chiplet Interconnect Express, or UCIe specification. I first wrote about it in the summary of SC22’s advancements and have since got even more excited about its promise after talking to its architects. So what is UCIe, and why am I holding it in such high esteem?
To answer this question, we need to first look at the problem that UCIe and other technologies like the Compute Express Link (CXL) are trying to solve. Moore’s Law is running out of gas, and shrinking process technology, the foundation for computing performance advancement since the birth of the microprocessor, will only get us so far in advancing semiconductor density. While CXL has provided a fantastic industry standard to connect chips on a computing platform and has eliminated historical proprietary interconnect schemes that have gated true heterogeneous computing, UCIe is going a leap further enabling this same industry standard foundation on chip package. What does this mean? Chiplet architectures, first envisioned by DARPA’s CHIPS program back in 2017, can finally be advanced without the limitation of proprietary technology. This means that both industry and consumers of computing can dial in the right balance of CPU, GPU, DPU etc chiplets into a a microprocessing package to deliver the best performance capabilities for targeted workloads. The who’s who of semis have engaged in the UCIe specifications both from a standpoint of silicon suppliers (enter AMD, ARM, Intel, NVIDIA Samsung, TSMC and more) and large service providers with their own silicon aspirations (hello Alibaba, Google, Meta, and Microsoft). Notably absent is Amazon, and we’ll need to watch to see if they really want to fight an standard consensus that seems destined to take off in products across the industry.
So why am I elevating UCIe to this lofty position? Microprocessor advancement is still foundational to breakthroughs of everything that runs on silicon. Without access to more performance our advances will be gated to today’s processing capabilities, and the computing industry sits alone in terms of the rate of innovation enjoyed for decades based on silicon advancement. Further breakthroughs simply require it, and for me that is a big call for applause for the team who delivered the UCIe 1.0 specification. Expect more details on the TechArena soon. As always, thanks for engaging – Allyson.
