Alessandro Perilli 🇺🇦

Meet Willow, our state-of-the-art quantum chip “Willow has state-of-the-art performance across a number of metrics, enabling two major achievements. - The first is that Willow can reduce errors exponentially as we scale up using more qubits. This cracks a key challenge in quantum error correction that the field has pursued for almost 30 years. - Second, Willow performed a standard benchmark computation in under five minutes that would take one of today’s fastest supercomputers 10 septillion (that is, 1025) years — a number that vastly exceeds the age of the Universe.” We live in a exciting time - advancements in quantum computing are incredibly transformational. But what happens when you combine two transformative technologies? That’s the potential between quantum and AI - opening up limitless opportunities.

8

Alan Turing's 1950 paper was not his final word on the subject of machine intelligence. Here is a reading of the text of a lecture he gave on the topic to the BBC audience in May of 1951. Turing's views on the topic continued to evolve during the early 1950's as he met frequently with the British Ratio Club ( https://lnkd.in/eDySmeqc ) , developing and first presenting the 1950 paper to the group while continuing to incorporate their feedback into his thinking. This 1951 BBC lecture was part of that evolution, which even included his earlier proposed Imitation Game (aka the Turing Test), as has been nicely explored by Andrew Hodges in his, "What Did Alan Turing Mean by 'Machine'?" Turing intended to get back to the topic of machine intelligence before his untimely death in 1954. As he mentioned in his reply to Claude Shannon and John McCarthy's 1953 invitation to contribute to their collection on machine intelligence, Automata Studies ( https://lnkd.in/ey-Tip77 ), he was declining because he had "been working on something quite different for the last two years: viz. the mathematics of morphogenesis, though I expect to get back to cybernetics very shortly." It is fascinating to consider where his brilliant mind would have gone next, and certainly worth revisiting what may have been his last public words on the topic. Enjoy! #turing #imitationgame #machine #intellgence https://lnkd.in/evEP36Pu

41

5 Comments

005: Quantum Gates - The Building Blocks of Quantum Circuits Quantum gates are the quantum equivalent of classical logic gates. They manipulate qubits through quantum operations, allowing us to create complex quantum algorithms. Some basic quantum gates include the Pauli-X, Pauli-Y, Pauli-Z, Hadamard (H), and CNOT gates. 🔍 Why are Quantum Gates important? Quantum gates are essential for constructing quantum circuits. They enable operations like superposition and entanglement, which are the core features that give quantum computers their unique computational power. Below is a simple example in Qiskit using the latest version to demonstrate some basic quantum gates. In this example, we create a quantum circuit with a single qubit. We apply a series of quantum gates: the Pauli-X gate flips the qubit state, the Hadamard gate creates a superposition, and the Pauli-Z gate applies a phase flip. Finally, we measure the qubit to observe the result. Stay tuned as I explore more quantum gates, their operations, and how they contribute to building powerful quantum algorithms. #QuantumComputing #Qiskit #QuantumGates #TechInnovation

3

The UK’s AI Action Plan: A Blueprint for Innovation & Growth Imagine this: £14bn in commitments from tech giants, 13,250 jobs created & AI Growth Zones revitalizing the UK economy. The UK government’s new AI Opportunities Action Plan is a game-changer for both innovation & economic growth. Here’s what you need to know: 1️⃣ Economic Growth and Infrastructure 🔹AI Growth Zones: Revitalising regions like Culham, Oxfordshire, with cutting-edge AI projects. 🔹Supercomputing Power: Major investments to turbocharge the UK’s AI capabilities. 2️⃣ AI in Public Services 🔹Streamlining admin tasks for teachers & public sector workers. 🔹AI-powered cameras to identify infrastructure repairs (goodbye potholes!). 🔹Advanced diagnostics in healthcare, including early cancer detection. 3️⃣ Private Sector Collaboration Tech giants like Vantage Data Centres, Kyndryl & Nscale are building hubs in Liverpool, Wales, & beyond. 4️⃣ Social Impact 🔹Future-proof jobs for de-industrialised regions. 🔹A National Data Library and AI Energy Council to manage data security & energy use. The Big Picture: Fully embracing AI could contribute £47bn pa to the UK economy over the next decade. But it’s not without challenges—ensuring ethical oversight, safety & learning from past AI missteps are critical. The UK is making its mark as a leader in the AI revolution. 🚀 Where do you see the biggest opportunities in this plan? P.S. Share this with your network ♻️ if you’re excited about the future of AI in the UK.

5

1 Comment

𝗖𝗵𝗮𝗻𝗴𝗲 𝗶𝗻 𝗺𝗼𝗯𝗶𝗹𝗶𝘁𝘆 𝗵𝗮𝘀 𝗮𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗲𝗱 𝘁𝗵𝗮𝗻𝗸𝘀 𝘁𝗼 𝗡𝗩𝗜𝗗𝗜𝗔'𝘀 𝗹𝗮𝘁𝗲𝘀𝘁 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻𝘀 𝗶𝗻 𝗔𝗜 𝗮𝗻𝗱 𝗰𝗼𝗺𝗽𝘂𝘁𝗶𝗻𝗴. After CES 2025, NVIDIA's new advancements are transforming the auto industry. Earlier, autonomous vehicles faced challenges with detection and response times. But due to NVIDIA's technology, including the GB10 superchip, safety and environmental awareness in cars have increased. 𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱? - Magnus Ostberg from Mercedes-Benz and Martin Kristensson from Volvo shared how NVIDIA's AI boosts vehicle compute power. - Jensen Huang unveiled the GB10, ready to redefine various AI applications and humanoid robots. - NVIDIA and Uber partnered to enhance autonomous vehicle development using Cosmos and DGX Cloud. The result was a commitment to efficiency and safety in transportation, with strategic partnerships making headlines. Now, Toyota, Aurora, and Continental are leveraging NVIDIA DRIVE AGX Orin for driverless trucking with DriveOS. 𝗧𝗵𝗲 𝗹𝗲𝘀𝘀𝗼𝗻? Integration of AI is rapidly reshaping mobility. Are we ready? #NVIDIAInnovation #AutonomousVehicles #AIFuture #MobilityTransformation #TechInAutomotive

1

With all due respect to The governing bodies of developed nations - regulations and laws are well intentioned, however without the proper control elements and functions they cannot maintain the insight, gain access to the metrics and telemetry needed to hold AI companies accountable. There are 4 core things needed to give AI regulations teeth - (1) agency over data with a denial of access core architecture to thwart data scraping and brokerage, (2) representation for data agents, subject of content/data so they have the ability to negotiate use of their content, (3) transaction management so data agents and owners can actively manage adherence to data terms and conditions, provenance and use by AI companies and hold them financially responsible for breach of contract, (4) sovereignty in digital space to ensure that individual rights can be upheld identically to their rights and privileges in the real world. There are many companies and organizations who are working towards these goals. We need support in order to stand up and provide solutions that allow for control.

2

Might be a major step forward #secure #communication using #quantumtechnology: "...entanglement of two nuclear spin memories through 40 km spools of low-loss fibre and a 35-km long fibre loop deployed in the Boston area urban environment." https://lnkd.in/ets4yD3K based on this Nature paper: https://lnkd.in/eQRa3J9q

3

#GenAI Blogpost by #founder Patrick Wendell and Mosaic founder and race car driver Naveen Rao. The evaluation of monolithic #AImodels to #compoundsystems is an active area of both academic and industry research. Recent results have found that “state-of-the-art #AI results are increasingly obtained by compound systems with multiple components, not just #monolithicmodels.” These findings are reinforced by what we see in our customer base. Take for example financial research firm FactSet – when they deployed a commercial #LLM for their Text-to-Financial-Formula use case, they could only get 55% accuracy in the generated formula, however, modularizing their model into a compound system allowed them to specialize each task and achieve 85% accuracy. #largelanguagemodels #genai https://lnkd.in/gQdDWNdR

14

Congrats to the Tabnine team!! - Gartner evaluated them across all 5 AI use cases in Gartner® Critical Capabilities for AI Code Assistants against 11 other vendors and ranked Tabnine #first for #CodeGeneration, #CodeDebugging, and #CodeExplanation use cases, and second for #CodeModernization and #ArtifactBuilding and #Testing use cases. The company also released the newest chapter of their O'Reilly guide on AI for the #SDLC. Customers are enjoying the choice between various #LLM's including #Claude 3.5 Sonnet and #GPT-4o in real-time. I love #investing in #extraordinary #founders - we're living in the Wow Era!

20

1 Comment

AI is navigating a very tricky transition from "novel" to "useful" at the moment, and while the wild creative swings and the (now-maligned) wearables are capturing the attention, there's a deeper discussion happening about what it means to get generative models ready for reliable, robust, enterprise-grade deployment. More in our breakdown of this week's top #fashiontechnology news: interli.net/4aYFUhn

38

4 Comments

What do Aravind Srinivas (CEO of Perplexity.ai), Yann Lecun (Chief AI scientist, Meta), Steve Case (founder of AOL), Dinesh Maheshwari (CTO Groq), and Ramin Hasani (CEO Liquid AI) have in common? They all participated in the Imagination in Action summit at Massachusetts Institute of Technology (MIT) yesterday. and this is just a short list of the amazing speakers and participants at this amazing event. Here are a few key takeaways: 1. Human relationships, human collaboration, and human innovation are going to be even more important in the age of AI. It was important for me to start with that, as many people believe AI is taking over the world and that is not the case. which leads me to: 2. The smartest people in the AI world today are concerned about the potential negative impacts of AI, and are investing significant resources to reduce those risks. That being said - they are running forward at an astonishing speed, so I am not sure what they are doing is enough. 3. Several breakthroughs are required in order to allow us to get to AGI, and to make this technology scalable to every process and every device. the current compute requirements are just to high. companies like Groq with their LPU innovative chip design that runs 10-1000 times more efficient than NVIDIA JPUs, and companies like Liquid AI with their innovative AI architecture, are spearheading some of these efforts. 4. There is a huge number of REALLY SMART people who are working on using AI to solve some of the biggest problems of the human race and our planet, such as curing cancer, fighting viruses, supporting the elderly, and much more. It was truly inspiring to see. 5. The entrepreneurial spirit is alive and well. over 100 startup founders participated in the event. 80 of them came to give a 90 seconds pitch (yes, 90 seconds). 6. Future models will have significantly less hallucinations, they would be able to understand our world (which they do not today), and hence, will be able to reason and plan. 7. While the number one concern about AI at the enterprise level is Data Security, there are more and more solutions that allow you to benefit from AI benefits while maintaining 100% control over your data. 8. The open source world of AI, is a critical component to the successful future of wide AI implementation 9. The future of code writing and interacting with computers is going to change dramatically. Natural language, voice, and vision, will allow us to program and use computers in a much more intuitive way. Computers will adapt to humans and not humans to computers 10. The conversations outside of the 4 presentation halls were as fascinating as the speakers and I met some amazing people, and I am grateful for meeting every one of them 🙏 BTW - I am still in Boston today (Friday the 19th), so if anyone wants to meet and talk, show me cool things in Boston or MIT, or geek about AI, or just chat, drop me a DM) Thank you John Werner for putting this amazing event together!

36

15 Comments

🚨BRILLIANT🚨 The #EU 🇪🇺 just built an #AI tool called 'Archibot' or 'Ask the EP Archives' that makes all the EP archives available through AI 👏 Using #Claude, their archives of about 2.1M documents are now (hopefully) easier to navigate with #GenAI We need more of experimental AI deployment like this! Obviously, this is one of the best ways to use LLMs - build tools like Archibot - that can help us search the EU archives... Something most of us are very excited to do... 🙄 Only the The EU AI Act consists of 180 recitals (introductory statements), 113 Articles, and multiple annexes detailing specific requirements and classifications 🤯 Clearly, #Archibot would be useful to go through that 🤷‍♂️ ▶ Check it out yourself: https://lnkd.in/dxBUfhnx ▶ Click the button that says 'Online access to Archival Documents' ▶ Click the 'Select dashboard' menu at the top right and choose 'Content analysis' ▶ Click 'Ask the EP Archives' tab on the main page (second tab) Enjoy! #llms #anthropic #languagemodels #artificialintelligence #generativeai #europe #europeanunion #archiving #euregulation #regulation #airegulation

7

#LLM based #Agents for #Software #Engineering: A Survey The recent advance in Large Language Models ( #LLMs ) has shaped a new paradigm of #AIagents . Compared to standalone LLMs, LLM-based agents substantially extend the versatility and expertise of LLMs by enhancing LLMs with the capabilities of perceiving and utilizing external resources and tools. To date, LLM-based agents have been applied and shown remarkable effectiveness in #SoftwareEngineering ( #SE ). The synergy between multiple agents and human interaction brings further promise in tackling complex real-world SE problems. In this paper, the authors collected 106 papers and categorize them from two perspectives, i.e., the SE and agent perspectives. They present a comprehensive and systematic survey on LLM-based agents for SE and also discuss open challenges and future directions in this critical domain. https://lnkd.in/gmy6RkV7

1

Blimey. This is a bit good. The lady in the video below does not exist. Microsoft has just released a paper introducing VASA; a framework that takes inputs of a single image and an audio clip and outputs a hyper-realistic talking face video with incredibly accurate lip-sync and lifelike facial gestures. It is important to note that this is a research project sharing results (there are no plans presently to make this a commercially available product or API), but it's still fascinating to see how far we've come in such a short space of time when it comes to AI-generated video (😧 I still get the shivers when I think about Will Smith eating spaghetti. If you know you know) source >>> https://lnkd.in/eJPn4Rrf

11