MIT/Conference/IIA/25/04/15
Business - April 2025 Summit Schedule
Pioneer Stage
9:00am Block 3 | Pioneering Scientific Super Intelligence to Rapidly Expand Human Knowledge
Panel - left to right
Riccardo Sabatini
- Title: Chief Data Scientist, Orionis Biosciences
- Background: Riccardo Sabatini is a physicist and data scientist best known for decoding the human genome and contributing to precision medicine. He has worked at the intersection of AI, biology, and data. At Orionis Biosciences, he focuses on using advanced computational biology and data modeling to discover new therapeutic pathways and accelerate drug development.
- My Notes
Person/Carina Hong
- Bio
- Title: CEO, AxiomAI
- Background: Carina Hong is a technology executive and founder focused on AI and data-centric solutions. As CEO of Axiom, she leads efforts to develop tools that use artificial intelligence to analyze complex scientific data and generate actionable insights, particularly in health and genomics.
- links
- My Notes
- "mathematical superintelligence"
- searching Chain of Thought for moments of Alethea or Inspiration and then training AI to seek that
- idea - if you just do the math (without any context), the creation of math tools can lead to new market tools. Babbage machine, quant finance, engineering simulations.
- automated Theorem Proof
Takeaways
- Bio
Person/Geoff von Maltzhan #CEO of lila.ai
- Science/Life
- Background: Geoff von Maltzhan is a biotechnology entrepreneur and General Partner at Flagship Pioneering. He is a founder of multiple pioneering biotech companies and holds numerous patents. His latest venture, Lila.AI, explores the intersection of artificial intelligence and life sciences, seeking to accelerate the discovery and development of therapeutics using AI-first approaches.
- My Notes
John Werner
- Title: Founder & CEO, Imagination in Action
- Background: John Werner is a serial entrepreneur, community builder, and catalyst known for launching initiatives that connect technology, media, and storytelling. He has deep ties to MIT Media Lab and is a frequent speaker and curator of innovation gatherings. His organization, Imagination in Action, creates forums where global thinkers share how emerging technologies can positively shape society.
My Notes
09:28 US/MA AI Strategic Task Force
9:30am Block 4 How State-of-the-Art Open-Source GenAI is Leveling the Playing Field for Every Business
- John Werner
- Person/Bob Young Role/Founder of Redhat Bob Young (businessman) - Wikipedia
- My Notes
- apparently his wife runs a "needlepoint" business and they are using AI for customer service - https://needlepoint.com/
- My Notes
- Person/Alvin Graylin
- Person/Karl Zhao for DeepSeek
- Karl Zhao, PhD - LinkedIn
- GEN AI Consultant
- DeepSeek AI
- Quasar Digital Solutions
- Sicore AI
- My Notes
- "when answers become cheep, the questions you ask and your ethics matter most" #Quote from Person/Alvin Graylin
- Person/Bob Young is "the father of open source" ? Open/Source
- Person/Karl Zhao
- Person/Alvin Graylin
- "universities were not making significant contributions to research" because you "had to be at a private lab like OpenAI" to make a difference, but DeepSeek and Open/Source or Open/Weights can enable University to participate again
- Person/Karl Zhao makes the argument that open source systems are more secure, even with respect to AI
10:00am Block 5 | AI Adoption: Success Stories and Implications of AI Agents
Panel - left to right
Person/Paul Baier
- Paul Baier | LinkedIn
- Title: CEO, Co-founder
- Organization: Company/GAI Insights
- Background: Paul Baier is a technology executive and entrepreneur focused on enterprise AI adoption and implementation strategies. At GAI Insights, he leads efforts to help organizations successfully integrate and scale AI solutions across their operations, with a particular focus on measuring and optimizing AI investments. He is also an Executive Fellow at Harvard Business School and Forbes Columnist.
- My Notes
Person/Lisa Huang
- Lisa Huang | LinkedIn
- Title: Head of AI Investment
- Organization: Company/Fidelity
- Background: Dr. Lisa Huang leads AI investment strategies at Fidelity, focusing on identifying and implementing AI solutions that enhance investment decision-making and client services. Her work spans across quantitative analysis, machine learning applications in portfolio management, and AI-driven market insights.
- My Notes
- Transcript
AI Journey at Fidelity
- Longstanding AI involvement: Lisa joined in 2018; Fidelity has had an AI center for years.
- Recent shift to incubation: Now leading efforts to ideate on next-gen AI innovations.
- General AI is a game-changer: Fidelity sees the emergence of generative AI as a timely opportunity to evolve quickly.
Key Challenges
- High cost of error: Mistakes in financial services can have billion-dollar consequences.
- Slow, thoughtful adoption: Compared to consumer tech, financial AI requires rigorous governance and risk controls.
- AI as augmentation, not replacement: Focus is on enhancing human capabilities, not replacing them.
“Take a human. That human has a certain value. Add AI to it.”
Person/Robert Sala
- Robert J Sala | LinkedIn
- robert sala image
- Title: CTO
- Organization: Company/Prudential Financial
- Background: Robert Sala serves as the Chief Technology Officer at Prudential Financial, where he oversees the company's technological transformation and AI integration initiatives. He leads strategic efforts to modernize financial services through AI adoption while maintaining robust security and compliance standards.
- My Notes
- they are using GitHub/CoPilot
- uses the term Data Gravity - new #Term to refer to when data is easier to use in a certain platform (for example, with AWS/Bedrock or Salesforce/Agentforce)
"architecturally, we have a principle to assume it's never more than a 24 month decision and there are always two way doors"
- Transcript
- Person/Robert Sala:
- When I talk to, you know, you, John, everything should be repatriated. When I talk to Amazon, everything should go to bedrock. You know, there's a continuum that we're all going to likely play, and I think we're all trying to figure out what the right topography is, for where, like, where should compute should live. There's a data gravity component of that. There's a whole bunch of things, right, cost of inference and watching that move around. When you think of the sprawl, we talked about this a little bit earlier, this reminds me, Dave, myself, this reminds me of the dot-com explosion, right? Think of all the website, site, shop-in-a-box kits and tools and platforms and people and online automated email forms and submissions and postings and content management, there were hundreds and hundreds of them. And what we saw over time is that it all settled down and they coalesced and there was consolidation and there was acquisition and sort of it grooved into a much more manageable entity. I could be wrong, but my theory is that we're going to see something similar here. Right now, it's such a wide open space with so many unknowns and everybody's racing into it hoping to be the, sort of, the hotter chair when the music stops, so to speak, and we all consolidate. You know, time will tell, but it is a little crazy right now. The only other thing I'd add is... Do you find yourself, RJ, going with just your incumbent vendors, let's just keep them housed, we don't want to do new ones for now, or do you market the startups out here? Yeah, architecturally we have a principle that says assume everything is no more than a 24-month decision and strive to make everything a two-way, right? Because today's amazing capability might get leapfrogged three times in the next 14 months and what we don't want to do is lock in. And so that's hard to do, but we spend a lot of time, not so much saying tool A, tool B, tool C, we spend a lot of time saying whichever tools we do, how do I press the clutch and swap it back out, not unless we know how to do it. Two quick things I would say is, one, I'd say focus on the problem to be solved and there are a lot of different ways and tools that you can solve it. So if you build to be agnostic to the technology...
- Person/Robert Sala:
Person/Karen Stroup
- karen stroup
- Karen Beck Stroup | LinkedIn
- Title: Chief Digital Officer
- Organization: Company/WEX
- Background: Karen Beck Stroup drives digital transformation initiatives at WEX, a global leader in financial technology solutions. She focuses on leveraging AI and emerging technologies to enhance payment solutions and customer experiences across corporate payment systems.
- My Notes
Person/Jieyu Fan
- Jieyu Fan - LinkedIn
- jieyu fan
- Title: Head of Advanced Analytics, US
- Organization: Company/John Hancock
- Background: Jieyu Fan leads advanced analytics initiatives at John Hancock, focusing on implementing AI and data science solutions in the insurance and financial services sector. Her work involves developing and deploying AI-driven solutions for risk assessment, customer service, and operational efficiency. She has been particularly active in developing cutting-edge AI and Gen AI solutions in the industry.
- My Notes
My Notes
How do you think about competitive differentiation?
- Person/Jieyu Fan - think of it like land
- what are the use cases that drive your organization?
- we are working for investors
- understanding their daily workflow and trying to marry the tech with the market
- "I used to not understand what Product/Market/Fit meant, I probably still don't"
- Person/Jieyu Fan - think of it like land
how do you think about open vs closed and behind the firewall vs exposed?
10:30 Block 6 🧠 The Future Internet of AI Agents and Why this is a New Frontier
- Aaron Pressman — Tech Reporter, Boston Globe
- Person/Ramesh Raskar — Associate Professor, MIT Media Lab
- Ramesh Raskar, MIT Media Lab, Camera Culture group, Cambridge MA, USA
- talks about MCP with his solution Person/Ramesh Raskar/GitHub/NANDA
- My Notes
- Transcript
- and navigate, and socialize, and learn, earn, and transact on our behalf, how are they all going to do this? You know, how does this look like the DNS, the browsers, the HTTP and HTML of the web? The exciting part is companies like Anthropic and Google have recently released protocols for communication. It's like the TCP IP of, you know, two machines talking to each other, which is great. But you have to go well beyond just communication and solve problems like discovery, search, and authentication, and trace. And that's what Nanda is about. It's an architecture that will support many protocols, including MCP and A2A, but each of those protocols could have dozens of different implementations. So that's the beauty of thinking like an architecture here. And so it requires all those things, just like, you know, a website gets certificates, there are coupons like weblogs, there's registration, all these things that we have in the web world also map to the agent world, but it would be a mistake to say that agents will behave just like the web pages behave, because in the 90s we could have said, oh, web pages will behave just like fax machines. So we have it. You know, we have the Nanda registry, like a DNS registry, and I'm delighted to announce that 15 AI universities are on this note already. Isn't that amazing? Remember, the very first nodes on the Internet were UCSB, UCLA, and SRI, and MIT came just right after that, so we're very delighted that we're including UCLA in this as well. And many companies, many startups already have their services on this open source, open architecture platform, and you can go play it out at nanda.mit.edu. And let me show you a quick video of how this works. Oh. Can we go back to the video? All right. What you'll see in the video is that to, you know, design this, to organize this birthday party, I can connect to various services, my Spotify list, you know, the bakery shop, you know, I can add all the services that I want. And they might be charging me. They might charge me, like, 50 cents for the query, you know, or do some intelligence on their sites. You can go to all these servers and add them. If there's a new shop in your neighborhood that shows up, you can just add them to your services. And this looks like a chat GPT or a cloud-like interface for today. You can play with it. And then you can go and say, hey, are these real stores? You know, there shouldn't be bad actors on this thing. And you can say, hey, you know, what's the weather, just to make sure everything is working. What tools do you have? Say, oh, I can, you know, I can make cakes for you. I can design songs for you. I can send party supplies for you, and so on. And all the services will be out there working for you. And just for a few dollars, just like you are willing to pay a few dollars to get, you know, the best airline tickets, because, you know, you are paying for it or order from Uber, some of those prices are included, all the services will be out there. So finally I can say for my daughter's birthday, hey, you know, do you have, you know, a chocolate cake? And can you deliver that by tomorrow? And it goes to the Nanda agent of the bakery shop and talks in real time. And there's a small intelligent model working on the bakery shop's server as well. And they can look at their inventory and their list and come back with an answer. Right? So all the orchestration around your life can be done this way. Of course, we are taking the low-stakes scenario, like a birthday party, but imagine doing the same thing for, you know, your health condition. You know, you have to go and search for all these pages, and all the services are coming at you, or think of a supply chain, and so on. So we are very delighted that we are launching this open source, and if your company and if your startups can get up and running in a matter of minutes with whatever existing software you have, you can add the Nanda modules on top of it. So, you know, you can post them.
AI Notes
initial notes
- Slide 1: “Towards the Web of Agents”
Slide 2: "The NANDA Spec: Enabling Agentic Ecosystem"
- This explains how NANDA (Networked Agents + Decentralized Auth) enables a decentralized, agent-based web through a set of components:
- Certificates
Identity verification from registry to client. - Coupons
For reputation, transactions, and discovery. - Registration
Client/server registration via the SDK. - SDK-specific schemes
Enables registry actions like search and reputation lookup from the NANDA SDK. - Registry Options
Flexibility in choosing the certifying registry.
GitHub link: github.com/aidecentralized/nanda-servers
- Here’s a clean transcript of Ramesh Raskar’s remarks, organized into logical sections and lightly edited for clarity:
Transcript: Ramesh Raskar on the Web of Agents and NANDA
“And navigate, and socialize, and learn, earn, and transact on our behalf — how are they all going to do this?”
We need to think about what this new world looks like. How does it resemble the DNS, browsers, HTTP, and HTML of the early web?
The exciting part is that companies like Anthropic and Google have recently released protocols for communication — the TCP/IP equivalent for AI agents talking to one another. That’s a great start, but we have to go well beyond communication to solve problems like:- Discovery
- Search
- Authentication
- Traceability
That’s what NANDA is about.
What is NANDA?
NANDA is an architecture that supports multiple protocols — including MCP and A2A — and each of those protocols can have dozens of different implementations. That’s the beauty of designing for extensibility.
Just like the web needs:- Certificates (identity/auth)
- Weblogs (logging and reputation)
- Registration (domain registry, etc.)
… we’ll need similar components for agents. But here’s the key insight:
“It would be a mistake to assume agents will behave just like web pages.”
In the 1990s, people might have said: “Oh, web pages are just like digital fax machines.” That turned out to be completely wrong — and so will underestimating agents.
NANDA in Action
We already have the NANDA registry, much like a DNS registry. And I’m delighted to announce that:
15 AI universities are already running nodes — including UCLA, just like the early days of the Internet when UCSB, UCLA, and SRI were among the first nodes.
Dozens of companies and startups are already onboarded, using this open-source, open-architecture platform.
You can try it yourself at: nanda.mit.edu
Use Case: Planning a Birthday Party with Agents
Let me show you a video (technical glitch: “Can we go back to the video?”)
Imagine you’re planning a birthday party:- You connect services like your Spotify playlist, your local bakery, or a party supply shop
- They might charge you small fees, like 50 cents per query
- You can interact with them through a ChatGPT-like interface
You might ask: - “Are these real stores?”
- “What’s the weather like tomorrow?”
- “What tools do you have?”
And you’ll get agent responses like: - “I can bake cakes for you.”
- “I can design songs.”
- “I can send party supplies.”
Now let’s say:
“For my daughter’s birthday, can you deliver a chocolate cake by tomorrow?”
The request goes to the NANDA agent at the bakery. A small AI model on their server checks their inventory and availability, then responds.
This orchestration is lightweight, modular, and real-time.
Beyond Birthday Parties
Birthday parties are low-stakes, but imagine this model applied to:- Healthcare (coordinating treatments, searching for services)
- Supply Chains
- Education
- Government services
Open Invitation to Developers
If you’re a startup or company, you can get started in minutes using your existing software stack. Just add NANDA modules on top.
It’s fully open source, and we invite you to try it out and host your own agents.
Let me know if you want this adapted into a one-pager, slide, or Logseq format.
10:45 Block 6.5
- CrusoeAI CrusoeAI
- Person/Nadav Eiron Nadav Eiron LinkedIn
- Person/Chase Lochmiller Chase Lochmiller | LinkedIn
- chase
- he's virtual
- My Notes
- Transcript
1 -
- So, come from, and what do you mean by vertically integrated AI infrastructure? Sure. So, you know, I think, you know, from my perspective, when I first started, you know, learning about and playing around with a lot of AI, and, you know, we sort of had this boom in deep learning after, you know, the release of AlexNet, you know, a lot of the outcomes really felt magical, right? It felt like, you know, humans had sort of, like, discovered a fundamentally new technology, and, you know, it's sort of this, you know, sort of these superpowers that, you know, weren't really imaginable or possible previously. And when I was thinking about, you know, what was driving that, it was really about the fact that we suddenly had, you know, or, sorry, let me take a step back. It wasn't a fundamentally new breakthrough in terms of, you know, the statistical methodologies that led to, you know, deep learning being possible. You know, multi-learning neural networks have existed for, you know, many decades. But what did make it possible was the fact that we suddenly had a lot more data just through the advent of the Internet, and we suddenly had enough computing power. We were able to parallelize that computing power through, you know, the use of GPUs to actually train, you know, these multi-learning neural networks with things like backpropagation to converge on solutions that were actually meaningful and useful. And, you know, I sort of saw this playing out, and, you know, I really felt like intelligence embedded in every aspect of the human experience was going to be, you know, a fundamental improvement for everyone and everything. Here, let me interrupt because I've got a lot of questions for you. Where are we in the race for AI infrastructure? How do you respond to critics or concerns of an AI overbuild? And do you have a customer yet? And maybe expand why you're particularly interesting these days and why. You almost got the Secretary of Energy to be on this panel, but that didn't work out. Go. Sure. So, you know, AI infrastructure at scale really leads to this convergence of energy and computing. You know, the computing infrastructure to run AI at a really meaningful scale just requires tremendous amounts of energy. So, you know, that's really sort of what led to the Bruce Hilda statements, energy first approach to building large-scale computing infrastructure, and why the Secretary of Energy is so interested now in AI is that energy is one of the key bottlenecks and key requirements to making AI happen at scale. We do have a customer. We do have many customers, both large and small. I think – I'm not sure if you're referring to our Avalon facility in terms of where we have a customer. We do have a customer there.
2 -
- If you're pessimistic, why do you think, you know, this is a doomed, what do you call it, doom loop, or a doomed, yeah. Yeah, I mean, let's see. I don't want to answer that question. So I'm always an optimist, and, you know, I think Crusoe's philosophy in all this is that, you know, we've taken an energy-first approach, which means actually bringing the demand for computing to areas that we can access low-cost, clean, abundant energy. You know, energy, you know, we can effectively produce clean energy at scale on a cost-effective basis. We can't do it everywhere because of transmission, distribution, you know, just naturally it doesn't always, you know, the sun doesn't always shine, the wind doesn't always blow in the areas where we actually need to use energy. So our philosophy is, you know, build data centers in areas where the wind is blowing and where the sun is shining, and where you can access low-cost, clean, geothermal energy like markets like Iceland, where we've built, you know, quite a bit as well. But West Texas is, you know, the best market in the United States for wind and solar, and that's really what led us to building in Abilene, Texas, is that we can access low-cost, clean, abundant wind and solar energy. Great. I have two more questions for you, and then I've got one for you.
3 -
- Sure. So, I mean, really what we're focused on trying to deliver is what I call a negative-grade premium for computing, which means that we can deliver a computing infrastructure that is powered both cleaner and cheaper than the existing infrastructure on the market side. The way we affect that is that, for the first time in history, we're actually seeing clean energy solutions being the cheaper ways to produce energy. Now, again, it's not in all cases and in all places, but we're able to deliver clean computing infrastructure by co-locating that infrastructure in areas where we can access that clean and low-cost energy. Great. So, Nadav, there's a lot of talk about the latest AI models and the race towards AGI, but you still see value in some of the older models. Tell us about that, what technologies are still needed to advance for broad AI collaboration, and what types of applications are businesses running on your platforms? Yeah. Hard to agree with that. And then, finally, I'm going to come back to you. I'll foreshadow this. Chase, what advice would you give other entrepreneurs? My business partner, Abe, references you all the time, saying you are awesome. I'm just curious what advice you have for entrepreneurs looking to start an AI company, so we'll come back to you. But, Nadav, let's see how you take this. So, first, Crusoe is a vertically integrated AI infrastructure company, and Chase talked a lot about the deep end of the vertical going lower in the stack towards the energy, the physical, the buildings, the locations. We're also vertically integrated up the stack, and I think one of the places AI needs to expand today is it needs to be easier to use for the people that are the ultimate users, I would say, in the business community. We talked about this creation of wealth. It needs to come through people that have businesses, successful businesses, being able to access technology, and we believe that technology is too hard to access today for those people. And so, that's when we talk about use cases and a bunch of work in creation. We're working towards, like, okay, how is it not about the latest and greatest model, but it is about how you take this model and implement it in your business. I was talking to someone earlier today, and the analogy came about is kind of like Amazon in the early days of the internet, right? Like, they're a retail company. That's not like a very traditional business, maybe one of the first businesses ever. And they were a huge force in pushing things together, because the technology was at a place where you could focus on retail, yet use the technology and benefit from it and make your business look different to the world. And so, I think we're at that moment in AI, and we want to make AI accessible to those industries, and those industries really care about how it helps their bottom line and not necessarily about the bleeding edge of the problem. Yeah. All right. So, Chase, are you on your sixth mountain? If you can let us know. And then, any advice? What year were you to graduate from MIT? I was class of 2007. Well, I was actually class of 2008. I graduated in 2007. And, you know, I grew up in Colorado around the mountains, and always had this passion for mountaineering and wanted to climb the highest peak on each continent. So, I've done five of the seven summits. I have two remaining. I'm not sure I'll get to them any time soon, because I'm quite busy building Rousseau and accelerating the proliferation of AI. So, a little known fact. Thomas Jefferson wanted the summer White House, or even the White House, to be out in Colorado, and it didn't work out. But I do wonder if that had happened, if your business wouldn't have been in Virginia, but maybe out there. All right. Advice to our founders here? Advice to entrepreneurs? You know, look, I think there's a couple pieces of advice I would say, is that, number one, no one gets their own. You know, life is pretty amazing in terms of there's a lot of really talented, amazing people out there that are often really willing to help you. And you haven't met these people, some of you have, some of you haven't. And if you're really passionate and ambitious, and you can paint a very clear vision for what the future might look like, there are people out there that will help you. And some of them are going to benefit, and some of them are just going to do it out of kindness and warmth. But, you know, there is an incredible dearth of people that are willing to help support you, and I see this huge worth invested in whatever crazy idea you have. So you guys, leaving you wanting more, Nadav is giving a lightning talk. So if you want to hear him give...
- AI Notes
1 -
- Transcript: Chase Lochmiller on AI Infrastructure and Energy
- Q: What do you mean by vertically integrated AI infrastructure? Where are we in the AI infrastructure race? And how do you respond to concerns about overbuilding?
- ⸻
- When I first started learning about and experimenting with AI — especially during the post-AlexNet deep learning boom — the outcomes honestly felt magical. It was like humans had discovered a new kind of superpower.
- But stepping back, what made that possible wasn’t some brand-new statistical breakthrough. Multilayer neural networks have been around for decades.
- What changed was:
• The explosion of data, thanks to the Internet.
• And the availability of computing power, especially parallelization via GPUs, which allowed us to train those networks effectively using backpropagation. - So the magic wasn’t new math — it was about scale, infrastructure, and integration.
- ⸻
- On Vertically Integrated AI Infrastructure
- At Crusoe, we believe the key to advancing AI infrastructure is a vertically integrated approach, where:
• Energy and compute come together in a single unified system.
• We optimize for both resource availability and sustainability. - This is critical because AI infrastructure at scale requires massive energy. That’s why there’s a convergence happening between energy policy and AI infrastructure.
- It’s also why the Secretary of Energy was nearly on this panel — the DOE is deeply interested in this space now, because:
- Energy is one of the primary bottlenecks for scaling AI.
- ⸻
- On Customers and Demand
- Yes — we do have customers. In fact, we have many, both large and small.
- One example is our Avalon facility, where we’ve deployed infrastructure with an active customer already onboard.
2 -
- Here’s a continuation of the transcript from Person/Chase Lochmiller during the Crusoe CEO conversation, focusing on energy-first AI infrastructure and optimism vs. doom loops:
Transcript: Chase Lochmiller on Optimism, Clean Energy, and Crusoe’s Strategy
Chase Lochmiller | LinkedIn
>
Q: If you're pessimistic, why do you think this is a doomed… doom loop?
I try not to be pessimistic — I’m always an optimist. At Crusoe, we’ve built our entire approach around this optimism with what we call an energy-first strategy.
That means:
>
Bring the demand for compute to where we already have clean, low-cost, abundant energy.
We can produce clean energy at scale and cost-effectively — but not everywhere. The challenge is transmission and distribution. The sun doesn’t always shine and the wind doesn’t always blow exactly where compute is needed.
So instead of trying to force energy to data centers, we bring data centers to energy.
On Location Strategy
We deploy infrastructure in places where:- Wind is blowing
- Sun is shining
- Geothermal energy is available
For example: - We’ve built quite a bit in Iceland — clean geothermal energy.
- West Texas (e.g., Abilene, TX) is arguably the best market in the U.S. for wind and solar. That’s why we chose to build there.
“We can access low-cost, clean, abundant wind and solar energy in West Texas — that’s the key.”
- Here’s a continuation of the transcript from Person/Chase Lochmiller during the Crusoe CEO conversation, focusing on energy-first AI infrastructure and optimism vs. doom loops:
- 3 -
- CrusoeAI CrusoeAI
E15 Basement, Bartos Theater
- 10:50 Zarathustra vs. Socrates: Genius and LLMs - Person/Peter Danenberg
- Peter Danenberg - LinkedIn
- Transcript
- AI Notes
11:00 Levels of AI for Scientific Discovery and Engineering - Person/Kushagra Tiwary
11:10 Collaborative Visual Intelligence with Person/Annika Thomas
- Home
- Hey! My name is Annika Thomas and I work with Jonathan How in the Aerospace Controls Laboratory. My research interests include robotic perception in multi-agent localization and mapping, computer vision applications for intelligent systems, and autonomy in space.
- Graduate Research Assistant in the STAR Laboratory at MIT
- My Notes
- Home
11:20 Person/Jascha Bach
- ChatGPT/Link 11:20 Jascha Bach and Consciousness
- Joscha Bach - Wikipedia
- My Notes
- could AI already be conscious?
- is there a secret ingredient that makes all the difference?
- he thinks it could
- the question is - is the strange loop itself, is it so complicated
- can we test for consciousness?
- nova
- TODO this guy is amazing, I need to read his stuff - do an intake
- LiquidAI building biological systems that are more useful. he mentions a talk at 7
11:45 Human AI Agents: Perspective-Awareness in Agentic AI Person/Hossein Rahnama #Amazing
- ChatGPT/Link 11:40 AI Perspective-Awareness Paradigm
- Hossein Rahnama | LinkedIn
- 11:40 AI Perspective-Awareness Paradigm
- My Notes
- Perspective-Aware AI PAi
- graph-based approach
- CAD drawings
- metadata
- sub-graphs are added to bigger graph object called a chronical
- new data objects called Chronicles
- ABMs - agent based modelings
- amazing but confusing
- how the heck to they detect seniority or competence from slack messages
- TODO post human documentary on bloomberg with emily chang videos with others - check this out!
- fine tuned mapping facial expressions
- can choose the chronicle to talk to - pose a question - graph will be posed - so it ends up creating a synthetic person - it's a model that not just the facial expression but the bottom left corner is an audit dashboard
- you can post a question to two chronicles and let them debate them to develop your opinions
- contextual household model generation - collective intelligence of a community in terms of energy, recycling
- it's early days
- 10:50 Zarathustra vs. Socrates: Genius and LLMs - Person/Peter Danenberg
- Atrium 3rd floor - part of
13:20 Fireside Chat: The Future of Work with AI
- 1:20 PM – 1:30 PM – 3rd Floor Atrium
Speakers:
• MacKenzie Sigalos – CNBC
Person/Ayush Chopra – Camera Culture - Person/Michael Casey – De/centralize/d/AI/Society DAIS
- Dave ...
- My Notes
- 1:20 PM – 1:30 PM – 3rd Floor Atrium
- E15 Basement, Bartos Theater
MIT/Conference/IIA/25/04/15/1345 Block A Agentic Futures Internet AI Agents
MIT/Conference/IIA/25/04/15/1400 Block B Lightning Talks
- VANA Vana Builder Workshop - MIT Decentralized AI - 60 mins 4:30pm ET - Google Slides
- chatgpt data export can be a proof of data ownership
- proof of uniqueness of data
- look in the comments to get to the templates in the speaker notes - proof of contribution python template - this is then all written on chain
- that proof is publicly available but the data is encrypted
- some Data/DAOs use client side ZK
- PrimeInsights generates a proof via a Chrome/Extension - PrimeInsights DAO
- you can contribute your coinbase exports or your trading history
- once you have proof of contribution
- VRC-20 is the VANA standard for ownership of dataset
- examples - MindDAO is for
- DNA DataDAO
- you can mine ownership of the overall VANA network
- you don't need anybody's permission to start a datadao
- if you want to start the spotify datadao, you can do that, because each user actually owns their data on spotify
- building the interface for the user to contribute
- today we do have to build the interface, but they are working on making a unified wallet
- metta mask
- knowhere
- getknowhere.xyz
- solidproject laid out this pod storage
- erbit was like users controlled compute storate - compute that's trusted by the user
- it basically runs client side
- where do you work with the unencrypted data ... in aggregate level it's in trusted environments
- most teams are at least 2-3 people and initial build takes 1 week to build a data dao
- if you are not a developer you actually can still create a data dao since
- this uses solidity
- dexa scans, medical records
- What DataDAOs to Build in 2025
- someone asks about what are the typical revenue amounts
- typically a company goes to 10k people that would sell imessage data, $10-50 and 60% failed to sell
- there's a fortune500 data company that buys spotify data, but if you pair it with fashion data then that's worth $25
- users are getting rewarded in this data token
- early on they were paying people in gift cards
- with the reddit data dao, people were getting paid with the square root of their karma
- if you had a lot of karma, you could get $400
- the token is volatile and you could
- someone asks how does bundling happen?
- is there more dollars
- where is the storage?
- is there visibility to the user of how the dao is used?