Humaina’s Post

Is AI more like a car or a search engine when it comes to safety? California's SB 1047 is sparking intense debate on this very question. The proposed legislation mandates that AI companies with significant investments in training "frontier models" conduct safety testing or face liability for catastrophic events. While critics argue this could stifle innovation, the conversation about AI safety and accountability is crucial. If developers truly believe their AI systems are safe, then compliance should not be burdensome. https://lnkd.in/gpcEZrJE

AI regulations have been "in the works" since OpenAI released its first ChatGPT model, with some leaders like Sam Altman publicly calling for more regulation in the industry, and undermining those that would impact their company in private. California's bill presents a unique case where the legislature is attempting to provide some kind of framework for ensuring AI systems are "safe" in the sense that they won't go rogue and cause a mass-casualty event. Predictably, there's a split amongst AI researchers, who aren't entirely sure how dangerous AI may actually be. Ultimately, this bill will likely be challenged and watered-down to the point that we won't see a meaningful impact on the industry or wider economy for some years, but it's telling that AI researchers who think the models are entirely safe and concerned about having to face new regulations. #ai #compliance #publicpolicy #bigtech #openai #regulation

TIL about constitutional AI (“Rather than use human feedback, the researchers present a set of principles (or “constitution”) and ask the model to revise its answers to prompts to comply with these principles.”) #llms #ethicalai

In 2021, several employees at artificial intelligence company OpenAI—the company that created the now-ubiquitous AI chatbot ChatGPT—left OpenAI to start their own AI company, Anthropic. Today, both companies claim to do cutting-edge AI work, and to care deeply about making AI beneficial for all. But a key concept on the minds of many policymakers and AI thought leaders is AI safety: the management of the novel risks that come with AI. As AI systems improve and embed themselves further into society, the need to manage the potentially catastrophic, societal-scale risks that AI could introduce will increase. It's unclear whether private interests such as OpenAI and Anthropic are the best candidates for doing crucial frontier work on AI safety, according to Vox senior correspondent Dylan Matthews. Check out Dylan's fascinating article for a thoughtful look into these companies' efforts to mitigate safety AI concerns: https://lnkd.in/d_FGpCEU Our upcoming program, “Futures Masterclass: Anticipatory Policy Innovation and Decision Making”, is all about decision-making in an uncertain and ever-changing world. We'll point you firmly in the direction of forward-thinking policymaking, and teach you to embed foresight practices into your organisation to strengthen its anticipatory capabilities. Sign up for the program here: https://lnkd.in/d8bufaGZ

In 2021, several employees at artificial intelligence company OpenAI—the company that created the now-ubiquitous AI chatbot ChatGPT—left OpenAI to start their own AI company, Anthropic. Today, both companies claim to do cutting-edge AI work, and to care deeply about making AI beneficial for all. But a key concept on the minds of many policymakers and AI thought leaders is AI safety: the management of the novel risks that come with AI. As AI systems improve and embed themselves further into society, the need to manage the potentially catastrophic, societal-scale risks that AI could introduce will increase. It's unclear whether private interests such as OpenAI and Anthropic are the best candidates for doing crucial frontier work on AI safety, according to Vox senior correspondent Dylan Matthews. Check out Dylan's fascinating article for a thoughtful look into these companies' efforts to mitigate safety AI concerns: https://lnkd.in/gQZr2Msv Our upcoming program, “Futures Masterclass: Anticipatory Policy Innovation and Decision Making”, is all about decision-making in an uncertain and ever-changing world. We'll point you firmly in the direction of forward-thinking policymaking, and teach you to embed foresight practices into your organisation to strengthen its anticipatory capabilities. Sign up for the program here: https://lnkd.in/d8bufaGZ

"If I build a car that is far more dangerous than other cars, don’t do any safety testing, release it, and it ultimately leads to people getting killed, I will probably be held liable and have to pay damages, if not criminal penalties. If I build a search engine that (unlike Google) has as the first result for “how can I commit a mass murder” detailed instructions on how best to carry out a spree killing, and someone uses my search engine and follows the instructions, I likely won’t be held liable, thanks largely to Section 230 of the Communications Decency Act of 1996. So here’s a question: Is an AI assistant more like a car, where we can expect manufacturers to do safety testing or be liable if they get people killed? Or is it more like a search engine? This is one of the questions animating the current raging discourse in tech over California’s SB 1047, legislation in the works that mandates that companies that spend more than $100 million on training a “frontier model” in AI — like the in-progress GPT-5 — do safety testing. Otherwise, they would be liable if their AI system leads to a “mass casualty event” of more than $500 million in damages in a single incident or set of closely linked incidents." https://lnkd.in/gV-_z9ye #AI #ResponsibleAI #SB1047

Where AI predictions go wrong

Where AI predictions go wrong

The debate over large language models (LLMs) and their potential to usher in artificial general intelligence (AGI) has reached a fever pitch. In a recent analysis, former OpenAI employee Leopold Aschenbrenner argues that we may be on the cusp of LLM-based general intelligence capable of any task a human remote worker can do. However, this viewpoint is met with skepticism from experts like Yann LeCun and Gary Marcus, who question whether scale alone can overcome the limitations of LLMs. While the future of LLMs remains uncertain, one thing is clear: the implications for AI policy and oversight are profound. https://lnkd.in/gVRi8uSa #AI #LLM #Patents #IP #VC #DeepTech

Where AI predictions go wrong