About
StandardC is the privacy-first, AI-powered platform for banks, credit unions, and other…
Articles by Robert
Activity
6K followers
-
Robert Mann shared thisTHE ALIEN IN THE VAULT Bill Gates told CNN's Fareed Zakaria GPS on Sunday that AI is "a new alien species that is going to be smarter than we are." Most banks are answering that alien with a policy memo. Six months ago, Mike Freiling, PhD (MIT AI Lab, advisor to StandardC) wrote that generative AI is a supernova, "...throwing off new capabilities and new risks almost every day." His conclusion: "The genAI supernova will continue to explode." Both agree on what matters for banks. Nobody is coming to govern this for you. Gates says society has not caught up to how powerful AI is. Freiling is blunter: "Guarantees of confidentiality by third-party vendors cannot be relied upon." The question for every bank CEO, CRO, and BSA officer: If an alien intelligence is already reading your customer files through a chatbot tab, who is watching it? Governing AI in a bank is not a ban. It is four things: 1. Private data is removed before any model sees it. 2. Numbers calculated by code, not generated. 3. Every finding is traced to its source. 4. A human decision on the record. Gates also argued that a monitoring layer over AI can be built without significantly slowing progress. Banks can build their solution this quarter with StandardC Privacy First AI. Read the full article before your next board meeting: Mike Freiling's risk framework and five questions every board should ask about AI. What is your institution doing about the alien in the vault? https://lnkd.in/gBCBNK57 #AIGovernance #Banking #CommunityBanks #CreditUnions #PrivacyFirstAI
-
Robert Mann shared thisLook closely at the second line of this image. The AI that generated it dropped a letter from "dressed." One letter. In a sentence about its own limits. It rendered the picture beautifully, got 99% of the words right, and delivered the result with total confidence. I kept it, because that is the whole argument in one frame: the error is small, plausible, and easy to miss, and nothing in the system knew it was there. The line belongs to Mark Smillie, author of The Feeling Machine, and he backs it with a patient. Elliot had surgery that cut the link between his emotions and his reasoning. IQ untouched. Memory untouched. And he could no longer make a decent decision. He could analyze every option and choose none of them well. That is a large language model. Brilliant, tireless, and constitutionally unable to tell what matters. As Mike Freiling, MIT AI PhD and CFA, put it on our webinar, machine intelligence is alien. It weighs everything equally because it has never lived through anything. It will hand you a 94% confidence score with the same composure whether it is looking at a loan file, a school, or a word it just misspelled. Now put that one-letter error in a debt service coverage ratio. So here is my question for banking executives and the investors backing AI in banking: why buy a system that is built to decide? We built StandardC AI on the opposite premise. Governed agents instead of prompts, so the institution's own policies define what the system reads, checks and cites. Deterministic code runs the ratios and reconciliations; the model only interprets. Private data is tokenized before the model sees a byte. Every finding cites its source and every run is reproducible. The system puts the application, the tax return, the P&L and the core deposit history side by side and shows a human where they disagree. The human, the one with a body, a career and a signature, decides. Bankers: ask your vendor who signs when the model is 94% sure and wrong. Investors: ask whether the company you are backing can survive that question in a regulator's office. Then watch the conversation that started this: https://lnkd.in/g7gVethZ #AI #Banking #AIGovernance #Fintech #VentureCapital #HumanJudgment
-
Robert Mann shared thisEveryone's racing to take humans out of the loop. We think that's the mistake of the decade. Today at 2 PM Pacific, StandardC's Robert Baron, Mark Smillie (author of The Feeling Machine), and Mike Freiling, PhD, get into why AI can out-calculate you but will never out-judge you, and what that costs you the moment the stakes are real. "Judgment Without Feeling? Human vs. Artificial Intelligence" Register: https://lnkd.in/g235sWPH #AI #ArtificialIntelligence #HumanInTheLoopWelcome! You are invited to join a webinar: Judgment Without Feeling? Human vs. Artificial Intelligence. After registering, you will receive a confirmation email about joining the webinar.Welcome! You are invited to join a webinar: Judgment Without Feeling? Human vs. Artificial Intelligence. After registering, you will receive a confirmation email about joining the webinar.
-
Robert Mann shared thisEveryone's racing to take humans out of the loop. We think that's the mistake of the decade. Today at 2 PM Pacific, StandardC's Robert Baron, Mark Smillie (author of The Feeling Machine), and Mike Freiling, PhD, get into why AI can out-calculate you but will never out-judge you, and what that costs you the moment the stakes are real. "Judgment Without Feeling? Human vs. Artificial Intelligence" Register: https://lnkd.in/g235sWPH #AI #ArtificialIntelligence #HumanInTheLoop
-
Robert Mann shared thisUnintentional AI sharing. Confident wrong numbers - AI hallucinations. Both are fixable, the one that is not: a generation that never learned to disagree with the machine. Join us on September 23 for a lively discussion about enhancing human judgment with AI.Robert Mann shared thisCan a machine have judgment without feeling? Join StandardC for a live fireside conversation with Mark Smillie, author of The Feeling Machine, and Mike Freiling, PhD, moderated by StandardC Chief Experience Officer Robert Baron. We will dig into where AI decisioning ends and human judgment begins, the dependency trap that erodes expertise, and what it all means for high-stakes regulated work. #ArtificialIntelligence #HumanInTheLoop #BSA #AML #Compliance #Fintech #AI
-
Robert Mann shared thisIs your company building AI that makes you smarter? Or obsolete? Mike Freiling was working in AI at MIT decades before most of us had even heard the term. I've known Mark Smillie since he was an investment banker in the early days. On September 23rd, they're going head-to-head on whether judgment survives without feeling, and what that means for anyone building or relying on AI.Robert Mann shared thisCan a machine have judgment without feeling? Join StandardC for a live fireside conversation with Mark Smillie, author of The Feeling Machine, and Mike Freiling, PhD, moderated by StandardC Chief Experience Officer Robert Baron. We will dig into where AI decisioning ends and human judgment begins, the dependency trap that erodes expertise, and what it all means for high-stakes regulated work. #ArtificialIntelligence #HumanInTheLoop #BSA #AML #Compliance #Fintech #AI
-
Robert Mann shared thisThere is a disorder loose in regulated industries, and the people who have it think they are being productive. Prompting Disorder: refining prompt after prompt to get consistent results from a system that samples the data instead of reading it, and agrees with you because it was trained to. Alone, it costs an afternoon and a token bill that nobody budgeted for. Inside a bank, it costs a methodology: five analysts, forty prompts, different ratings for the same customer, no record of who asked what. Version control dies quietly. Confidence in every prior rating dies with it. That is a finding waiting for an examiner. We solved it, and the cure is boring on purpose. Open the article. At a minimum, learn about the disease before it infects your next audit.You are absolutely right, and that is the problem.You are absolutely right, and that is the problem.
-
Robert Mann shared thisMy co-founder Richard Laiderman wrote the essay below about a Go match, and it ends up explaining most of what worries me about AI in banking. His framework is simple. When you are ahead, you reduce risk to protect your lead. When you are behind, you gamble, because a Hail Mary is your best remaining chance. Sound strategy on a Go board. Now look at how financial institutions are adopting AI. These are companies that are ahead: real customers, clean exam histories, regulator relationships built over decades. Yet many are deploying AI like desperate players. Prompts that give a different answer every run. Decisions no one can trace back to evidence. Confidential customer data handed raw to models that may train on it. Ask yourself: would your auditor accept an analysis you cannot reproduce? An institution with a lead should never take Hail Mary risk to keep it. The winning move is the simplifying one: AI where the same inputs always produce the same outputs, where every finding cites its evidence, and where PII is redacted before a model ever sees it. That is the architecture we built StandardC on, and it is why our banking clients treat AI as a moat, not a bet. Richard's full essay is worth your time. It covers Go, game theory, and why the bubble has not popped. Read it below.
Robert Mann replied to a comment
1w
John Sabo precisely. This hive small firms a new advantage in competing with larger organizations. This can open new markets with limited resources and elevates administrative to informed decisionnmaking.
-
Robert Mann liked thisTHE ALIEN IN THE VAULT Bill Gates told CNN's Fareed Zakaria GPS on Sunday that AI is "a new alien species that is going to be smarter than we are." Most banks are answering that alien with a policy memo. Six months ago, Mike Freiling, PhD (MIT AI Lab, advisor to StandardC) wrote that generative AI is a supernova, "...throwing off new capabilities and new risks almost every day." His conclusion: "The genAI supernova will continue to explode." Both agree on what matters for banks. Nobody is coming to govern this for you. Gates says society has not caught up to how powerful AI is. Freiling is blunter: "Guarantees of confidentiality by third-party vendors cannot be relied upon." The question for every bank CEO, CRO, and BSA officer: If an alien intelligence is already reading your customer files through a chatbot tab, who is watching it? Governing AI in a bank is not a ban. It is four things: 1. Private data is removed before any model sees it. 2. Numbers calculated by code, not generated. 3. Every finding is traced to its source. 4. A human decision on the record. Gates also argued that a monitoring layer over AI can be built without significantly slowing progress. Banks can build their solution this quarter with StandardC Privacy First AI. Read the full article before your next board meeting: Mike Freiling's risk framework and five questions every board should ask about AI. What is your institution doing about the alien in the vault? https://lnkd.in/gBCBNK57 #AIGovernance #Banking #CommunityBanks #CreditUnions #PrivacyFirstAI
-
Robert Mann liked thisScientists have uncovered new clues about the origins of the universe. For generations, asking what preceded the Big Bang was considered a purely philosophical puzzle rather than a scientific inquiry. Because space and time both originated with the Big Bang roughly 13.8 billion years ago, traditional physics suggested there was simply no such thing as 'before.' However, modern cosmologists are finding clever, indirect ways to peer back through the cosmic curtain. By studying subtle fluctuations in the cosmic microwave background radiation and searching for primordial gravitational waves, researchers are beginning to test several highly sophisticated mathematical models of our cosmic prehistory. Among the most prominent theories is the 'no-boundary proposal,' which suggests the universe has no starting edge and that time behaves like space at the cosmic poles, rendering a 'before' meaningless. Alternatively, theories of loop quantum gravity propose a 'Big Bounce,' imagining our universe was born when a previous one collapsed and rebounded. A third intriguing hypothesis envisions a CPT-symmetric 'mirror universe' existing on the other side of the Big Bang, where time flows in reverse. Ultimately, while we may never look directly into this deep past, new technologies are being developed that allow us to final begin testing these theories. And the shape of our current and future cosmos may very well holds the keys to understanding whether we arose from a bounce, a mirror, or a timeless quantum beginning. source: Scoles, S. (2026). What Came Before the Big Bang? Scientific American.
-
Robert Mann liked thisRobert Mann liked thisWhat happens when AI starts learning before the dataset exists? The training data can begin to emerge while learning is already underway. Most AI training follows a familiar sequence: Collect data → build a dataset → train the model The dataset exists first. But self-play opens up a different possibility. One model generates examples. Another model learns from them. As the learner improves, the generator keeps searching for examples near the edge of what the learner can currently handle. Too easy, and there is little to learn. Too difficult, and the example is not useful yet. So the useful training data changes as the learner changes. Generate → learn → generate the next useful challenge → learn again Recent work on Self-Play Pretraining with Zero Data started both the generator and learner from random initialization, without training either one on natural data. The surprising part: the models improved even on datasets they had never been trained on. "Zero data" does not mean learning without examples. The examples are generated during training instead of coming from a natural dataset prepared beforehand. Why does this matter? Useful training data can be expensive, finite, or difficult to anticipate in advance. 📚 High-quality data takes time to collect and curate. 💰 Creating more of it can be expensive. 🎯 What the model needs next can change as its capabilities improve. 🔄 A fixed dataset does not adapt when the learner does. Self-play changes that relationship. Instead of preparing every example in advance, the generator can keep searching for examples useful at the learner’s current level. If this direction matures, it could matter where useful examples are rare, expensive, or hard to anticipate: 🏥 Healthcare: progressively harder simulated cases. 💳 Fraud detection: new fraud patterns and edge cases. 🔐 Cybersecurity: new attack and defense scenarios. 🤖 Robotics: increasingly difficult environments and edge cases. This is different from repeatedly training on old AI-generated outputs. The goal is not to reproduce more of the same data. The generator is actively searching for new examples near the frontier of what the learner can currently handle. This does not mean real-world data goes away. But the architectural idea is important: Training data does not always have to be a fixed input prepared before learning begins. In some systems, generating the next useful training example could become part of the learning process itself. 👍 Like if you found this valuable. 🔔 Follow Surya Gutta for insights on building production AI systems. #SelfPlay #AITraining #SyntheticData #GenerativeAI #AIArchitecture
-
Robert Mann liked thisEveryone's racing to take humans out of the loop. We think that's the mistake of the decade. Today at 2 PM Pacific, StandardC's Robert Baron, Mark Smillie (author of The Feeling Machine), and Mike Freiling, PhD, get into why AI can out-calculate you but will never out-judge you, and what that costs you the moment the stakes are real. "Judgment Without Feeling? Human vs. Artificial Intelligence" Register: https://lnkd.in/g235sWPH #AI #ArtificialIntelligence #HumanInTheLoop
-
Robert Mann liked thisEveryone's racing to take humans out of the loop. We think that's the mistake of the decade. Today at 2 PM Pacific, StandardC's Robert Baron, Mark Smillie (author of The Feeling Machine), and Mike Freiling, PhD, get into why AI can out-calculate you but will never out-judge you, and what that costs you the moment the stakes are real. "Judgment Without Feeling? Human vs. Artificial Intelligence" Register: https://lnkd.in/g235sWPH #AI #ArtificialIntelligence #HumanInTheLoopWelcome! You are invited to join a webinar: Judgment Without Feeling? Human vs. Artificial Intelligence. After registering, you will receive a confirmation email about joining the webinar.Welcome! You are invited to join a webinar: Judgment Without Feeling? Human vs. Artificial Intelligence. After registering, you will receive a confirmation email about joining the webinar.
Experience & Education
Organizations
-
Angel Faces
Board Member
- Present
Recommendations received
3 people have recommended Robert
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Robert Mann in United States
2138 others named Robert Mann in United States are on LinkedIn
See others named Robert Mann