Impact of AI Development

The effort to protect innovation and open source continues. I believe we’re all better off if anyone can carry out basic AI research and share their innovations. Right now, I’m deeply concerned about California's proposed law SB-1047. There are many things wrong with this bill, but I’d like to focus here on just one: It defines an unreasonable “hazardous capability” designation that may make builders of large AI models liable if someone uses their models to do something that exceeds the bill’s definition of harm (such as causing $500 million in damage). That is practically impossible for any AI builder to ensure. If the bill is passed, it will stifle AI model builders, especially open source developers. Some AI applications, for example in healthcare, are risky. But as I wrote previously, regulators should regulate applications rather than technology.  - Technology refers to tools that can be applied in many ways to solve various problems. - Applications are specific implementations of technologies designed to meet particular customer needs. For example, an electric motor is a technology. When we put it in a blender, an electric vehicle, dialysis machine, or guided bomb, it becomes an application. Imagine if we passed laws saying, if anyone uses a motor in a harmful way, the motor manufacturer is liable. Motor makers would either shut down or make motors so tiny as to be useless for most applications. If we pass such a law, sure, we might stop people from building guided bombs, but we’d also lose blenders, electric vehicles, and dialysis machines. In contrast, if we look at specific applications, like blenders, we can more rationally assess risks and figure out how to make sure they’re safe, and even ban classes of applications, like certain types of munitions. Safety is a property of the application, not a property of the technology (or model), as Arvind Narayanan and Sayash Kapoor have pointed out. Whether a blender is a safe one can’t be determined by examining the electric motor. A similar argument holds for AI.  SB-1047 doesn’t account for this distinction. It ignores the reality that the number of beneficial uses of AI models is, like electric motors, vastly greater than the number of harmful ones. But, just as no one knows how to build a motor that can’t be adapted to harmful applcations. For open models, there’s no known defense to fine-tuning to remove RLHF alignment. And jailbreaking work has shown that even closed-source, proprietary models can be prompted harmful responses. Indeed, the sharp-witted Pliny the Prompter regularly tweets about jailbreaks for closed models. Kudos also to Anthropic’s Cem Anil and collaborators for publishing their work on many-shot jailbreaking, an attack seems hard to defend against. I hope you will speak out against SB-1047 it if you get a chance to do so. [Original text (with links): https://lnkd.in/gtn4H7YK ]

AI is no longer just decorating rooms. It’s redesigning how we live. AI can now rethink rooms, floors, and entire layouts—turning bold ideas into build-ready designs. Would you do floor like that? The data behind the shift: • 30–50% faster design cycles using generative layout tools • 100+ layout permutations generated from a single brief • Up to 20–30% improvement in space utilization • 10–25% energy savings when airflow, lighting, and thermal paths are simulated early • 40% fewer late-stage design changes thanks to digital testing What’s fundamentally different? AI treats floor plans like software systems: Pedestrian movement is simulated before construction Natural light and ventilation are optimized virtually Furniture, walls, and utilities are stress-tested digitally Cost, carbon footprint, and materials are optimized in parallel This enables: Smaller homes that feel larger Offices designed around productivity and wellbeing Buildings that adapt over time instead of aging poorly The biggest myth? AI replaces architects and designers. Reality: AI handles complexity and permutations. Humans focus on vision, culture, emotion, and identity. The future of architecture isn’t just smart. It’s generative, data-driven, and human-centric. #AI #Architecture #Design via @Visual Spaces Lab #PropTech #GenerativeAI #FutureOfLiving #SmartBuildings #Innovation

As we reach the midpoint of 2026, one thing has become increasingly clear: the conversation around AI has shifted. Again.    A year ago, much of the discussion centered on what's possible. Today, the focus is on what's practical: how we build the data foundations, governance, infrastructure, and workforce capabilities needed to create lasting value.    A few key learnings we're applying as we head into the second half of the year:  --AI transformation is ultimately a people transformation. Technology is advancing quickly, but realizing its potential means building a workforce that knows how to use it thoughtfully.  --AI fluency is a core capability. That means thoughtfully redesigning roles and workflows and giving employees the confidence to incorporate AI into their daily work while maintaining human judgment and accountability.  --The role of technology leaders is no longer focused only on delivering systems. We're helping shape enterprise strategy, building workforce capabilities, and creating the conditions for innovation to scale responsibly.    The pace of change isn't slowing down, but some principles remain constant: start with the problem you're trying to solve, invest in your people as much as your platforms, build governance from the beginning, and remember that every technology decision should ultimately create value for the people we serve.

To the public, Microsoft uses its reputation as an AI & sustainability leader to tell a compelling story: AI will do wonders to help solve the climate crisis. To fossil-fuel firms, Microsoft has a different message: AI will help them drill, baby, drill. For more than a year, I’ve been poring over hundreds of pages of internal Microsoft documents, many of which were shared with the SEC, and interviewing current and former employees and execs on the giant's engagements with the oil & gas (O&G) industry. Microsoft doesn’t just passively provide its services to these companies. It develops bespoke AI-enhanced tools for them, which it also markets to them as for the explicit purpose of optimizing and automating drilling, and maximizing fossil-fuel production. Here’s an example: In a slide deck from Jan 2022, Microsoft prepared an analysis that said its tools could allow ExxonMobil to increase its annual revenue by 0.8%, or $1.4 billion—$600 million of which would come from optimizing its drilling. Internal Microsoft employee reports estimate the O&G industry represents a market opportunity of between $35 billion to $75 billion - especially notable with the pressures the tech giant faces to continually show payoffs to its massive AI investments. In the last year, Microsoft has sought to leverage genAI hype to land more contracts. In Sept 2023, company execs noted on a conference call with more than 200 employees that the energy industry was turning to Microsoft in a way that had perhaps “never happened before.” It needed to “maximize this opportunity” & “lay out the pathway” to genAI. One such pathway? Using generative algorithms to model oil and gas reservoirs and maximize their extraction. Employees themselves have campaigned relentlessly within Microsoft to point out how the company talks out of both sides of its mouth. But Microsoft execs have seemed only to learn into it further: promoting AI for fossil fuel extraction behind closed doors while publicly talking louder than ever about the climate benefits of AI as well as the company's sustainability leadership. They told me that Microsoft's O&G engagements show the unsavory reality of how the company’s AI investments are actually used. Driving sustainability forward? Maybe. Digging up fossil fuels? As execs put it in that September conference call, it’s a “game changer.” My latest investigation for The Atlantic. Gift link here: https://lnkd.in/gyEUvxQ4

AI is shaping who gets hired, and how much they’re paid 💸 . A new peer-reviewed study from Germany reveals something most news outlets overlook: AI hiring tools often reproduce historical bias, and media coverage isn’t calling it out. What the study found: – Women are recommended lower salaries – Minority applicants receive fewer negotiation signals – "Neutral" algorithms reflect old-school discrimination – Accountability is rare. Intersectionality is ignored. – Regulation? Still lagging behind. Can de-biasing fix it? The study says no; not without policy, transparency, and oversight. 👉 I pulled key findings into a quick visual breakdown (swipe through the carousel) Have you seen these patterns in hiring or HR tech? Drop your experience or perspective below - especially if you work in tech, policy, or talent. 📄 Link to full study in the comments. #FutureOfWork #ResponsibleAI #HiringBias

Last night I got the latest newsletter from Casey Newton at Platformer (check them out - essential reading for anyone working in Tech), and they were discussing California’s AI Bill, rapidly making its way through the Californian legislature (SB-1047, also known as the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act). Given that the EU’s AI Act has just entered into force - I thought it would be a useful exercise to compare their approaches to AI Regulation. Both pieces of legislation aim to ensure that AI systems are developed and deployed responsibly, but they take notably different approaches in terms of scope, regulatory mechanisms, and the thresholds they set for what constitutes a high-risk AI model. SB-1047 is a targeted effort to regulate "frontier" AI models. These are AI systems that involve substantial computational resources—specifically, those requiring more than 10^26 floating-point operations per second (FLOPs) or exceeding $100 million in development costs. The AI Act sets AI Models which post “systemic risk” at 10^25 FLOPS. It’s interesting that the Californians have introduced a development cost threshold. This could become out of date extremely fast due to GPU improvements and efficiencies. SB-1047 is primarily concerned with the potential for these advanced models to cause "critical harm," which includes threats such as mass casualties, large-scale cyberattacks, or significant disruptions to societal infrastructure. Key provisions of SB-1047 include the requirement for AI developers to notify the government before training any model that meets these thresholds. It also mandates the implementation of stringent safety and security protocols, including a kill switch that can shut down a model if it is deemed dangerous. The bill creates the Frontier Model Division, a regulatory body tasked with overseeing compliance and enforcing these rules, including conducting third-party audits and mandating prompt reporting of any AI safety incidents. In contrast, the AI Act imposes stringent requirements on AI systems considered to pose systemic risks by exceeding the FLOPs threshold, including mandatory risk management, transparency, human oversight, and third-party conformity assessments. Additionally, providers of these systems must adhere to approved codes of practice or demonstrate alternative compliance methods to ensure safety and mitigate potential harms. Both SB-1047 and the AI Act diverge in how penalties are structured and enforced. SB-1047 allows for significant penalties, including fines based on the computing costs used to train AI models. The bill also provides for injunctive relief, including the potential shutdown of AI systems that pose a risk of causing critical harm. The AI Act, on the other hand, imposes substantial fines for non-compliance, with penalties reaching up to €35 million or 7% of global turnover for severe breaches, particularly those involving prohibited AI systems.

How AI Destroys Institutions. This study, by two professors of the Boston University School of Law, examines the impact of artificial intelligence (AI) on societal institutions, particularly focusing on how AI systems can undermine the foundational structures that govern social interactions and promote stability, justice, and prosperity. It identifies several detrimental effects of AI on institutions: the degradation of expertise, the disruption of decision-making, and the erosion of social bonds, and concludes that AI poses significant risks to the structural integrity of institutions that are essential for a just society. To mitigate these risks, the authors suggest that future research should focus on developing frameworks that integrate human oversight in AI systems, promote transparency, and reinforce the importance of human expertise in decision-making processes. In addition, fostering community engagement and social capital is crucial to counteract the isolating effects of AI technologies https://lnkd.in/ehuqRtYn

If you want to understand the AI future, watch an 18-year-old live their life. As the mom of an 18-year-old, I’ve spent years worrying about screen time, algorithms, and the blurry line between “real life” and “online life.” But lately, I’ve started to notice something else. My daughter isn’t toggling between two spaces. She’s living inside one continuously blended world. She’ll FaceTime a friend while walking to the café. She’ll scan a QR code on a thrifted jacket, find the designer’s Instagram, scroll their stories, and share it with friends before she’s at the register. She’ll sketch on her iPad with a livestream concert playing in the background, sending a voice note that sparks a plan that turns into a night out. For her, these aren’t separate steps. It’s flow. It’s reality. And it’s a live preview of how AI will reshape our experience of the world. Many of us still treat digital tools and AI like places we “go” to. But that model is already fading. What is emerging is something ambient, blended, and instinctive. Here is what that looks like and why it matters. AI will be ambient, not separate. We grew up opening browsers and clicking icons. Even months ago, most of us treated AI as a tool. That is changing. AI now lives underneath everything, surfacing suggestions, anticipating needs, and stitching together data. Think less “open ChatGPT” and more “it is already shaping your inbox, calendar, photos, and feed before you ask.” Co-creation is becoming the default. My daughter does not just consume content. She layers, edits, and remixes it instinctively. That is the shift AI accelerates. We will brainstorm, design, write, and analyze with it offering ideas and filling gaps in the background. This is not automation. It is extension. Context is replacing the interface. She does not “launch” tools. She moves, and technology adapts. AI will increasingly respond to location, tone, gesture, and data streams. The shift is from “I click” to “it knows.” Culture is leading the change. Teenagers are rewriting norms through behavior. Their habits pull technology forward. Adoption will not happen in one big moment but through millions of small shifts until the old ways quietly disappear. Fluidity matters more than literacy. We were taught to learn technology. They move with it. That fluidity will define future readiness. This shift changes how we experience the world, and it changes us. It is not inherently bad. It is simply different. Watching my daughter live this way is not just a parenting moment. It is a front-row seat to the next era of human experience. Maybe this is not so bad. Maybe it is exactly the lens we need to understand what is coming.

𝐃𝐞𝐠𝐫𝐞𝐞𝐬 𝐎𝐮𝐭, 𝐃𝐢𝐬𝐜𝐫𝐢𝐦𝐢𝐧𝐚𝐭𝐢𝐨𝐧 𝐈𝐧: 𝐓𝐡𝐞 𝐃𝐚𝐫𝐤 𝐒𝐢𝐝𝐞 𝐨𝐟 𝐀𝐈 𝐚𝐧𝐝 𝐒𝐤𝐢𝐥𝐥𝐬-𝐁𝐚𝐬𝐞𝐝 𝐇𝐢𝐫𝐢𝐧𝐠 We celebrated the fall of degree requirements. We welcomed skills-based hiring with open arms. We trusted AI to make recruitment fairer. But here's the uncomfortable truth: 𝐖𝐞 𝐦𝐚𝐲 𝐡𝐚𝐯𝐞 𝐭𝐫𝐚𝐝𝐞𝐝 𝐨𝐧𝐞 𝐤𝐢𝐧𝐝 𝐨𝐟 𝐛𝐢𝐚𝐬 𝐟𝐨𝐫 𝐚𝐧𝐨𝐭𝐡𝐞𝐫. Skills-based hiring is meant to 𝑙𝑒𝑣𝑒𝑙 𝑡ℎ𝑒 𝑝𝑙𝑎𝑦𝑖𝑛𝑔 𝑓𝑖𝑒𝑙𝑑. Yet, the tools powering it—AI resume screeners, video interview analyzers, "𝑐𝑢𝑙𝑡𝑢𝑟𝑒 𝑓𝑖𝑡" algorithms—are often trained on biased historical data. The result? ➡️ Ageism coded into filters. ➡️ Racial bias hidden in name-matching. ➡️ Neurodivergent candidates penalized by automated “personality” scores. Amazon scrapped its AI hiring tool when it penalized resumes with the word “women.” Workday faces lawsuits over alleged AI discrimination against Black, disabled, and older applicants. And many job seekers are ghosted—rejected by machines before a human ever reads their name. 𝐓𝐡𝐢𝐬 𝐢𝐬𝐧’𝐭 𝐣𝐮𝐬𝐭 𝐚 𝐭𝐞𝐜𝐡 𝐟𝐥𝐚𝐰. 𝐈𝐭’𝐬 𝐚 𝐭𝐫𝐮𝐬𝐭 𝐢𝐬𝐬𝐮𝐞. AI can’t be the future of hiring until we make it accountable, transparent, and human-centric. 𝐖𝐞 𝐦𝐮𝐬𝐭 𝐝𝐞𝐦𝐚𝐧𝐝: Diverse, inclusive training data Human-in-the-loop decision-making Regular audits of AI tools Legal and ethical oversight Innovation without ethics is just automation of injustice. Let’s not replace gatekeeping with ghostwriting—by robots. Do you think AI is helping or hurting fairness in hiring today? Share your thoughts. Follow Samichi Saluja for more bold takes on AI, job search strategy, and the future of work. #AIHiring #SkillsBasedHiring #RecruitmentBias #FutureOfWork #DiversityandInclusion #HRTech #ResponsibleAI #HiringFairness