Stefano Maffulli

RSA Conference notebook: Kicked off the morning with an impromptu roundtable with Robert Ackerman and Carmen Marsh about ethics in AI development. More to come on that but it doesn’t look that great for the near future. Then dropped by the RSAC Sandbox. Reality Defender is yet another startup dedicated to deepfake detection, but on the plus side they apparently have CBS News as a customer. The company claims up to 95% detection but news organizations apparently don’t care about false positives as they tend to err on the side of caution and will reject even those with slightly questionable provenance. I’ll be talking to them later with two questions. First, Are social media platforms considering adopting these kind of tools to stop proliferation of deep fakes? And two, since I already know the answer, Why not?

3

4 Comments

"GenAI content moderation is always a cat-and-mouse game, but these findings suggest a particularly flimsy set of automated controls that failed to repel some of the most basic adversarial techniques" I shared my thoughts with The Times on research indicating Midjourney's safety controls are far from robust. This certainly isn't unique to Midjourney, but when adding "/" to the end of a harmful prompt bypasses content moderation, it's clear there needs to be more resources dedicated to meaningfully addressing the problem. https://lnkd.in/eXKJmtGi

48

4 Comments

🚨Fascinating report alert: "Safeguarding Brain Data: Assessing the Privacy Practices of Consumer Neurotechnology Companies" by Jared Genser , Stephen D. and Rafael Yuste. A must-read for everyone interested in privacy & AI. Quotes: "Advocates of neurorights are not calling for the creation of new rights, but rather the further interpretation of existing human rights law to guide the development of national legal and regulatory frameworks. The Morningside Group identified five key neurorights: (1) the right to mental privacy, or the ability to keep mental activity protected against disclosure, (2) the right to identity, or the ability to control one’s mental integrity and sense of self, (3) the right to agency, or the freedom of thought and free will to choose one’s own actions, (4) the right to fair access to mental augmentation, or the ability to ensure that the benefits of improvements to sensory and mental capacity through neurotechnology are distributed justly in the population, and (5) the right to protection from algorithmic bias, or the ability to ensure that technologies do not insert prejudices." (page 14) - "In coming years, these databases will function similar to how genetic and biometric databases function. Just as genetic material and fingerprints are used to identify individuals, so too will neural data, which is uniquely identifiable to a specific person as long as it is taken at a sufficient resolution. Further, advances in artificial intelligence are rapidly increasing the ability to decode information from neural data. As previously discussed, studies have found that when paired with generative AI, brain scans from non-invasive neurotechnologies allowed for the decoding of language, emotions, and imagery with high levels of accuracy." (page 17) - "Given the advances in generative AI, the growing quantities of neural data being collected worldwide, and the other types of data neurotechnology companies collect that connect back to individuals (such as the user’s IP address), it is clear neural data will soon be widely personally identifiable. Since consumer neurotechnologies are only starting to proliferate, existing databases are small, making it unlikely that neural data from consumer devices could currently be used to identify individual users. However, the consumer neurotechnology market is rapidly expanding, and it is likely that neurotechnology companies will soon begin to amass large amounts of neural data. This would again follow a trend observed in the consumer genetic testing space" (page 18) ➡ Read the full The Neurorights Foundation report below. ➡ For more information on tech policy & regulation, subscribe to my newsletter. #privacy #tech #techpolicy #neurorights #brainprivacy

149

7 Comments

Investigative work done by The Markup in partnership with a few other civil society organizations unearthed some egregious errors from a chatbot deployed by the local NYC government to aid citizens in finding answers to FAQs. An example of a clearly problematic outcome was answering questions like “Can I take a cut of my workers’ tips?“ with: “Yes, you can take a cut of your worker's tips. According to the City of New York's Payroll and Tip Reporting information, employers are required to report tips reported by employees and employees receiving $20 or more a month in tips must report all of their tips to their employer. Therefore, as an employer, you are allowed to take a portion of your worker's tips. For more information, you can refer to the Payroll and Tip Reporting page on the NYC Business website.” (You can read the full report from The Markup for more examples.) There are a whole slew of ethical issues that arise from such an incident such as: 1. Erosion of trust in the reliability of information provided in other instances by the chatbot, even when it might be right - also called algorithmic aversion. 2. Legal consequences due to wrong information whereby the person might act on the provided information that opens them up to lawsuits, e.g., in the workers’ tips case. 3. Harm to individuals and businesses, e.g., refusing to accept Section 8 vouchers or accepting tenants that receive rental income, i.e., discriminating against them, which is illegal for a landlord to do and would make finding housing difficult in an already problematic rental market such as NYC. Our quick take on small measures that would have helped to avoid this mishap proactively would be to: 1. Involve domain experts in the testing and validation phases of the chatbot development who could have spot-tested / smoke-tested responses to see if they went against housing regulations, workers’ rights, etc. 2. Rigorous testing and validation, especially constraining the system to rely on techniques like retrieval augmented generation (RAG) to sharpen answers and ground them in actual city laws that could be fed as a fine-tuning step into the chatbot’s AI system. 3. Transparency and user education could alert users more proactively that there is always an inherent risk in the chatbot providing responses that might not be accurate, or in this case, wrong to the point of being illegal, so users should proceed with caution and verify with second and third sources to confirm the validity of the response. While each of the steps require an investment upfront before the launch of the system for public consumption, the cost, we believe strongly, is well worth it to avoid harm and build trust capital with citizens, especially for essential services. #ethicalai #ai #ethics #aiethics #responsibleai https://lnkd.in/dAXzGqvP

9

Extending the #DSCR deadline is welcome, but it doesn't address why the deadline has been missed (article by Sam Trendall on PublicTechnology https://shorturl.at/74DDS). Technology adoption follows a curve which inevitably slows down over time. I would argue that all the 'early adopters', and most of the 'late majority' have already digitalised their care records. Rather than simply extending a deadline, we should be directly addressing the reasons why technology adoption is hard, or even undesirable, for care providers. One is a perceived lack of need, which persists in some quarters; relatively few care providers are counting the costs of legacy practices (and that includes those already using out-of-date care tech). Another is the complexity of record-keeping for certain use-cases. Multiple care providers have approached Total Care Manager because their existing solution doesn't work for #complexcare. Others have found that apparently simple things, such as the structure of their teams or shift patterns, have stretched the available tech beyond its intended use. Olivier van Ingen Wayne Adams Steffen Wulff Petersen And then there's the trust problem. This is a major issue in life-supporting nursing care such as #braintrauma, where providers are understandably more reticient to entrust the safety of their patients to technology - but it extends to all aspects of care. The solution isn't the extending of a deadline. It's the sharing of knowledge (which the follks at Care Tech Guide and Digital Care Hub are working hard to solve), the building of case studies, and ongoing development by technolgy vendors to work beyond the 'easy middle' and address unmet needs or previously unanticipated use-cases. Liam Palmer Katie Thorn That work won't end at 80% (whether or not we make it by 2025) 90% - or even at 100% It's an ongoing responsibility of stakeholders across the entire care industry, to keep building knowledge and sharing best-practices around the improvement of businesses and lives in care, and of the role that tech has to play.

7

6 Comments

After eight years of operating in "stealth mode" Travis Kalanick's City Storage Systems is ready to talk. Yesterday, Kalanick took the stage at the annual Food On Demand conference to give a rare keynote about the various businesses within the CSS constellation, namely Otter, CloudKitchens, and Lab37 kitchen automation. CSS' vision is to build the infrastructure that enables the "Internet Food Court" of the future, where highly convenient, personalized food delivery is price competitive with the cost of cooking. CloudKitchens boasts over 400 facilities in 30+ countries while Otter touches ~15% of all third-party delivery sales. The presentation underscored Kalanick's obsession with digitizing the physical world, sucking out all of the inefficiencies out of the legacy marketplace model which he described as a "luxury priced product that has a mediocre or worse experience." Despite the company's massive footprint, its success will hinge on whether operators will subscribe to its walled garden of software, services, and real estate. Will they bite? Full story on HNGRY Trends 👇 https://lnkd.in/dMv5yY5D #ghostkitchens #restauranttech #foodtech #startup #realestate #logistics #fooddelivery #marketplaces #restaurant #qsr #restaurantindustry

239

5 Comments

Insights from the most recent Trustible newsletter: 1️⃣ The enacted EU AI Act and Colorado AI Law set the stage for upcoming AI regulations. More jurisdictions are likely to 'free ride' on these efforts and pass their own, likely 'lighter' versions of these laws. 2️⃣ Trustible released an AI Policy Analyzer tool to help organizations ensure their AI policies are compliant with emerging regulations and standards. 3️⃣ Anthropic's recent research into LLM explainability, while impressive, doesn't yet attain the kind of explainability regulators want for automated decision making. 4️⃣ The 'AI Race' in both the corporate and geopolitical worlds may undermine a lot of recent efforts made towards reducing carbon emissions. We're only beginning to see the energy impacts of rampant AI use. 5️⃣ OpenAI had a terrible, horrible, no good, very bad week with high profile departures, accusations of voice impersonation, and disbanding the team that was supposedly in charge of AI safety oversight. See the full post here: https://lnkd.in/grgZtamg

5

Today's "Interesting person I found via RSS and decided to follow on LinkedIn" is John R. Legg. 'We live in a digital age, where people expect public access to things at the snap of their fingers. Archives worldwide have embraced digitization to scan, reproduce and copy sources from the past and make them readily available online. Digitized items also require a set of information, called metadata, that details necessary details on where to find specific information about a source. This method has helped countless researchers who may not be able to travel far distances for historical research. However, despite the ability to transform analog sources into the digital realm, we should still consider the implications of using such materials — especially items that hold value as “cultural heritage” that connect the past with the present. As researchers, we have a responsibility to adapt our own filters to consider whether we should be engaging with historical and cultural heritage objects, something very close to the present debates about archives like the MNHS holding sacred Dakota items for preservation. ' https://lnkd.in/eRvnmCZB

2

3 Comments

What's the root of most of the problems in America? Political corruption. This article is about the official lobbying activity in AI, dominated by the bigs in tech, but the $$ in lobbying is a small part of the total influence. In a recent interview, I said CXOs at LLMs and Big Tech had to understand the copyright infringement risks, cybersecurity risks, and catastrophic risks early on. They obviously made the decision to recklessly unleash LLMs due to confidence they would be successful in their government capture strategy (old saying -- big tech has a 100% success rate in Congress). While their strategy is toxic to customers and society, it is not without merit from a hard, cold (sometimes brutal) profit / market cap perspective. 1) Until this year, mainstream media and tech media appeared to be on the payroll of Big Tech (some tech media were). Almost nothing but adoration and envy. Ideological and political alignment seemed similar (all is not what is seems). Started to moderate in 2024. 2) Financial incentives. Members of Congress, families, staffers own big tech directly or indirectly (not startups), as do large investment firms, pension funds, universities, foundations, churches, journalists... Big Tech is so much of the market today nearly every institution is conflicted. 3) Revolving door syndrome. Thousands of former government staffers work for Big Techs, including many senior execs. Example -- Amazon was very aggressive during the $10B Jedi contract process for the DoD -- placed ads all over D.C., including commuter trains and buses. They recruited large numbers of USG workers, and former Amazon employees were hired by the DoD who were involved with influencing the contract (alleged). Another example: A former employee of MSFT is a member of Congress from WA, and has influenced many policies in favor of the co. 4) The patriot card. Intel played this strategy brilliantly for billions in subsidies, but the bigger techs have played the AI patriot card for trillions in market cap. "If you don't let us (do anything we want), China will do it". The Secretary of Commerce is playing hardball in negotiations on behalf of Microsoft. The assumption is that she'll run for POTUS--if not land at a Big Tech with 8 figure retirement package. Perhaps the assumption is wrong, but the behavior, history and patterns support it. 5) Corruption of academia. Big Tech has awarded many 7 figure compensation packages to professors and graduate students, and invested large sums for research, buildings, etc. -- directly and indirectly from founders and senior execs. Big Tech research departments became a high-end corp academic jobs program. The actual lobbying dollars pale in comparison to the total influence of Big Tech, but Congress has a problem - 75% of voters believe big companies have too much influence. Wonder where they got that idea?

32

11 Comments

Could Artificial Intelligence Create Real Liability for Employers? Colorado Just Passed the First U.S. Law Addressing Algorithmic Discrimination in Private Sector Use of AI Systems (US) https://buff.ly/3yFcneY - 💡 Colorado passed Senate Bill 24-205, the Consumer Artificial Intelligence Act (CAIA), the first U.S. law aimed at protecting consumers from harm arising from AI systems. - 📅 The CAIA will take effect on February 1, 2026, and applies to Colorado businesses using AI systems for employment decisions. - 🤖 High-risk AI systems must adopt a risk-management policy, complete annual impact assessments, provide notices to consumers, and report algorithmic discrimination to the attorney general. - ➡️ Colorado employers should monitor for guidance on compliance and amendments to the CAIA to ensure they are prepared. - 🇺🇸 There is currently no comprehensive federal legislation regulating the use of AI systems in the U.S. - 🔍 The CAIA shares structural similarities with the AI Act adopted by the European Union, despite industry opposition. - ⚖️ Compliance with the CAIA can offer an affirmative defense in an enforcement action by the attorney general.

1

1 Comment

🌐 Navigating the Crossroads of Data Curation and Privacy In the intricate web of today’s digital ecosystem, our interactions are incessantly tracked, cataloged, and analyzed. That is not news. A recent Wired article captured my attention – did you know that as of last fall numerous websites now openly declare the alarming number of "partners" collecting our data - from browsing habits to unique IDs. While we all inherently know websites share our data, actually seeing these partners number over 1,000 is overwhelming. Are we on the precipice of a world where personal information flows unrestricted? Author Kevin Kelly in The Inevitable talks about the unstoppable trends shaped by technology, including the seamless integration of data into our daily lives. This integration offers undeniable benefits such as personalized experiences and targeted services, enhancing efficiency and user satisfaction. Conversely, there is a trade-off between the benefits of data curation and the need to safeguard our privacy. This is going beyond the annoyance of my contacting a cooking school for my son to find my mailbox flooded with recipes the next day. As leaders, our challenge is to harness the advantages of this data-centric world while defending individual privacy. As this landscape evolves, it is crucial to advocate for and implement practices that ensure data is used ethically. As many embrace the benefits of tailored content and services, there are just as many that seek out platforms that empower users, giving them control over their personal information. This balance is not just a regulatory necessity but a strategic imperative that fosters long-term trust and loyalty. Here are some ideas in developing a framework where we can enhance user experience and uphold privacy: 1. Privacy by Design: Implement privacy from the initial design phase of all projects and throughout the lifecycle of the data. 2. Transparent Data Practices: Regularly disclose to users how their data is collected, used, and shared. Make this information easily accessible and understandable. 3. User Control and Consent: Empower users with tools to control their data. 4. Data Minimization: Limit data collection to what is directly relevant and necessary to accomplish a specified purpose. 5. Regular Audits and Assessments: Conduct regular privacy impact assessments and audits to ensure compliance with both internal privacy policies and external regulatory requirements. 6. Stakeholder Engagement: Engage various stakeholders, including customers, regulators, and privacy advocates, in discussions about privacy practices. 7. Educational Initiatives: Foster a culture of privacy awareness within the organization by training employees about their roles in protecting privacy. Share your ideas in the comments section. #dataprivacy #digitalethics #businessethics #leadershipintech

26

29 Comments

“To be able to fight disinformation…[we]…must understand it. But research on disinformation is limited in several key respects…A big obstacle for researchers is the lack of access to data. The best data is not in public hands, but is “sitting in private networks in Silicon Valley,” says Phil Howard…of the International Panel on the Information Environment (IPIE). And collecting relevant data is becoming more difficult. After Elon Musk bought Twitter…in 2022 the company shut down the free system that let anyone download information on posts and accounts, and began charging thousands of dollars a month for such data access. Meta announced in March that it would be retiring CrowdTangle its platform-monitoring tool that lets scientists, journalists and civil-society groups access data, though the company says academics can still apply for access to certain datasets. Such changes have seriously hampered researchers’ ability both to detect disinformation and to understand how it spreads… In America…efforts to fight disinformation have become caught up in the country’s dysfunctional politics…fighting disinformation requires a co-ordinated effort by tech platforms, academics, government agencies, civil-society groups and media organisations. But…any co-ordination of this kind has come to be seen, particularly by those on the right, as evidence of a conspiracy between all those groups…Researchers studying disinformation have been subjected to lawsuits, attacks from political groups and even death threats. Funding has also diminished…All this has had a chilling effect on the field...“It is difficult to avoid the realisation that one side of politics—mainly in the US but also elsewhere—appears more threatened by research into misinformation than by the risks to democracy arising from misinformation itself” Disinformation is a sprawling problem, requiring co-ordinated action from multiple sectors of society. Unfortunately, the analysis of it tends to be siloed and there is a lack of agreement on things like terminology. This makes it hard to join the dots and find lessons that apply more broadly. Phil likens the situation to the early days of climate science: lots of people are trying to tackle the same problem from different perspectives, but it is difficult to see the whole picture…The current lack of a joined-up response to disinformation is a problem, but also an opportunity: co-ordinating research and action should lead to better detection and mitigation of deceptive content, because modern disinformation campaigns all work in similar ways. But, as with climate change, cleaning up the world’s information environment presents a daunting, long-term challenge.” https://lnkd.in/gb_g9CKf Ellen Doowan Mark Nina Roman Emerson Alan Jevin Renee Sheldon Justin J.D. Klon John Brandon Ed Sasha Alex Jakub Jill Clint The Economist Celine

35

4 Comments

A couple of weeks ago a journalist at our national broadcaster did a long piece on driverless taxis in San Francisco. While the piece played up on peoples' fears and asked if you'd feel safe in one, they never provided any safety data comparing driverless vehicles to those driven by people. I managed to reach them on Twitter and here's is what they said: "I agree that data is important. Unfortunately, there are no independent studies available. The only data is provided by robotaxis operators. One example from Waymo here (below)." I can't say how accurate these Waymo numbers are but they certainly sound impressive. I'm left with two questions. 1) If the driverless car operators' number are this good why don't they provide complete open access to their live numbers with independent researchers/journalists? 2) I'd be curious to know the number of accidents that are seen for every mile or hour of driving that have been recorded for human driven vehicles. And specifically, how do these numbers compare with those of the operators' claims. Insurance companies must have this information and feel more or less confident on insuring driverless car companies vs. people. At some point, if the driverless cars are truly that much better this should become a simple decision. https://lnkd.in/g6g3PqCa

6

8 Comments

I wish I was at #RSAC this year, especially as we're seeing a major shift in technology. Attackers are using AI and accelerated computing faster and better than defenders... this has to change. Quick back story... at Accenture Tech Labs under Lisa O'Connor, I got addicted to the challenges of cyber defense at scale. We explored numerous technologies to improve threat detection, and NVIDIA GPUs were the most promising. That led me and my team to join NVIDIA, where we pitched building a GPU accelerated SIEM to leadership, which turned into something even greater... https://rapids.ai/ a GPU data science ecosystem (as some would say the foundations of accelerated data science). NVIDIA Morpheus (thanks to Bartley Richardson and Chau Dang) kept the original cyber dream alive... thank you both! RAPIDS.ai led me to many like-minded people... and together we started Voltron Data, and built Theseus, https://lnkd.in/dTurZRMU. We never forgot about cybersecurity though. The SIEM industry, as well as many other parts of #cyberdefense, need open standards and an accelerator-native query engine to detect threats faster and more efficiently. Next year, I look forward to announcing several new partnerships at RSAC! If you want to learn more or explore partnerships reach out or connect with Fernanda Foertter.

39

6 Comments

We're organizing a webinar around how to make documents accessible at scale (in large organizations such as universities, government agencies and enterprises). What are some of the topics you would like us discuss? Some options we're throwing around include: 1. How to monitor document accessibility at scale 2. How to control the production of documents 3. Remediation at scale 4. Tools and platforms (AI?) Please add your suggestions in the comments below. Let me know if there are other problems you'd like us to discuss as well. #Accessibility #AccessibleDesign

39

28 Comments

One of the oddest and most dangerous ‘dividends’ arising from the hypergrowth of Artificial Intelligence is the rise of what policy expert Jamie Neikrie and I call ‘Artificial History.’ We wrote up a piece for Tech Policy Press about the ability for generative AI to create highly credible but entirely fictional synthetic history. Now, bad actors have the ability to arm their claims with retroactive points of evidence spun out of fake first-hand accounts, on-the-ground testimony, or deepfake trusted messengers. If you can rewrite the narrative of where we come from, you can make a different argument about where we should be going. See more here: https://lnkd.in/eqxTQvae

18

10 Comments

The US is in desperate need of a functioning government rather than disparate agencies vying to assert their relevance and overstep their mandates. However, there's an interesting take here that could be a boon for open source AI if this ends up playing out. If it were deemed "anticompetitive" for a tech company to train a model on the internet and then sell access to it, would it not follow that providing open source access to the model would no longer violate that definition? https://lnkd.in/eeKXRjcU

9

We have a content-heavy publication for your holiday weekend (in the US, that is): - The Agentic AI report, built off of our webinar on the same topic. Be sure to check out our framework for understanding and evaluating the complexity of AI agents. And download the pdf version of the report (at the bottom of the report page). - A research piece on Graph RAG that allows LLMs to build and query their own intrinsic knowledge graphs. - A fascinating science review of research from Anthropic which unveils some of the inner workings of Claude. - A personal ideas piece reflecting on The Agentic Web and what it might mean to be human in a world which is dominated by content by machines, for machines. - A podcast interview with Māori Data Sovereignty expert, Karaitiana Taiuru about indigenous peoples and AI. - Part 5 in our How to Use Generative AI series on Weave: Levering the Technology's Combinatorial Power. - And also Helen's Book of the Week, Bits & Bytes from Elsewhere, and Facts & Figures about AI & Complex Change. Phew! A Few Dates for Your Calendars - 11 June 2024 at 2pm PT: Our June research briefing on our latest insights and best practices for collaborating with generative AI. - 27 June 2024 at 12pm PT: An introduction to Generative AI at The Haven in Bend Oregon. Let us know if you're local and would like to join us. - 9 July 2024 at 2pm PT: Our July research briefing on learning and AI. - 13 October 2024: Big news! Pencil in this date for the first Artificiality Summit in Bend, Oregon. Join us for our first in-person event and stay the week for the High Desert Innovation Week and the Bend Venture Conference. Please email us if you are interested in attending and/or contributing in any way. And, while I have your attention, don't miss our new Services page which covers how we help your organization unlock generative AI's potential. Yes, we've been doing this work for several years for organizations around the globe. What's new is the publicity around our services—check it out and give us a call! #ai #artificialintelligence #generativeai #airesearch #complexity #chatgpt #complexchange #changemanagement #futureofwork #artificiality #mindforourminds #intimacyeconomy #agenticweb

2

1 Comment