Good morning, and happy Friday! We’ve got major headlines, two deep dives, and some Friday Facts to kick off the weekend.
We’re again looking at the uncertain future of U.S. AI regulation. While some states are pushing forward with their own restrictive AI laws, the federal government still lacks a unified approach. With Trump now president-elect, the direction of national AI policy could shift—will his administration bring clarity or deepen the divide?
In other news, AI is making waves in quantum computing, and Wendy’s is adopting AI to streamline supply chain and inventory management.
Inside Today’s Edition:
Top Stories 🗞️
AI Poised to Outshine Quantum Computing in Simulation 🤖
AI Regulation Stalled, but Path Forward Emerges 🚦
[FF Original] How Good Is It? 👾
[ICYMI] Interview with Saturnin Pugnet, Visionary Founder of World ID 📽️
Tools Transforming Workflows, Insights, and Media 🧰
🤔 FRIDAY FACTS
How Much Energy Does a ChatGPT Query Use?
Stick around for the answer! 👇️
🤖 COMPUTING
Why AI Could Ultimately Eat Quantum Computing’s Lunch
The Recap: Rapid advancements in AI-driven simulation are posing a real challenge to quantum computing’s anticipated role in complex fields like chemistry and materials science. While quantum computing holds theoretical advantages for certain problems, AI is showing it may reach useful milestones much sooner.
Highlights:
New neural network models have made AI the leading tool for simulating complex materials and quantum effects, particularly in weakly correlated systems.
AI-powered simulations, built on classical tools like density functional theory, can model systems of up to 100,000 atoms, handling most practical chemistry and materials problems with ease.
Companies like Meta and DeepMind have made AI models that excel in material discovery and modeling, challenging the need for near-term quantum applications.
AI approaches pioneered by researchers like EPFL’s Carleo are proving effective for highly complex quantum systems, typically the domain of quantum computing.
High costs, complexity, and a need for millions of qubits mean large-scale quantum computers are likely still decades away from surpassing classical or AI-driven simulations.
AI models require massive data for training, often generated through costly simulations, creating a resource barrier for many research teams.
Alongside AI, other classical methods for quantum simulation continue to advance, suggesting a competitive landscape rather than a clear AI or quantum dominance.
Forward Future Takeaways: As AI methods become more powerful and accessible, they could fulfill a significant portion of the tasks once reserved for quantum computing. This rapid AI progress may prompt quantum research to refocus on a narrower set of complex applications, like high-temperature superconductors, where quantum retains unique advantages. In the long run, AI’s gains might reduce the urgency for large-scale quantum machines, reshaping expectations and investments in both fields. → Read the full article here.
👾 FORWARD FUTURE ORIGINAL
The Art & Science of AI Evaluation
In the previous article, we introduced the notion of how evaluating LLM outputs is distinct from testing standard software output, because unlike the latter, the former is not deterministic. Let’s now take an end-to-end example to illustrate this, integrating other concepts we’ve covered in this series.
Let’s take the enterprise situation of a clothes store implementing a product-enquiry chatbot for use by its (potential) customers. The chatbot pops up on the store website and can answer all sorts of questions in relation to the product catalog.
Let’s see how the generative AI concepts we touched upon come into action in such a scenario.
Firstly, there is the question of product classification (or taxonomy) within the catalog - this is something we were used to, even before AI came along. You might have waded through a taxonomy tree such as mens > summer > casual, or womens > winter > travel.
This can be seen as part of the deterministic setup, and is typically stored in a conventional database (in other words structured data) as part of the conventional software setup. Nothing new there per se. Where it gets interesting, of course, is when this can be opened up for users to query in natural language, which is where our LLM ideas come in with a bang. → Continue reading here.
🚦 POLICY
U.S. AI Regulation Stalled by Industry Resistance, But Experts See a Path Forward
The Recap: The U.S. continues to grapple with AI regulation, with a mix of state-level initiatives and federal actions, but comprehensive national policy remains elusive. Despite challenges, experts see potential in existing laws and rising momentum for a unified approach that could drive meaningful AI governance.
Highlights:
Tennessee protected voice artists against AI cloning, and Colorado implemented a risk-based AI policy approach.
Although California passed various AI-related bills, Governor Newsom vetoed SB 1047, a significant transparency and safety regulation bill, after industry opposition.
The FTC has cracked down on data misuse in AI, and the FCC proposed rules requiring disclosures for AI-generated political ads.
The AI Executive Order established the U.S. AI Safety Institute, which assesses AI risks in collaboration with top AI labs, though its existence depends on executive support.
Tech leaders and investors like Meta’s Yann LeCun and Khosla Ventures' Vinod Khosla have actively opposed tighter AI regulations, citing concerns over hindering innovation.
With nearly 700 state-level AI bills introduced this year, pressure is building for a coherent national framework to avoid a fragmented regulatory landscape.
UC Berkeley’s Jessica Newman and California State Senator Scott Wiener believe the challenges have set the stage for future regulatory efforts, with increasing calls for federal policy.
Forward Future Takeaways: While the path to cohesive AI regulation in the U.S. faces hurdles, the urgency for oversight is growing. As AI’s potential risks and societal impacts become more apparent, industry resistance may soften in favor of stability and clarity in regulation. With substantial public and legislative interest, federal policy could emerge, steering AI development in a safer, more consistent direction. → Read the full article here.
🛰️ NEWS
Looking Forward: More Headlines
AMD Launches 1B-Parameter Model: AMD's open-source OLMo model offers strong reasoning and multi-task abilities, optimized for Instinct GPUs and Ryzen AI.
Grounding LLMs for Science: The paper presents a method to enhance LLMs in scientific tasks by adapting tool usage for improved accuracy.
Apple Explores Smart Glasses: Apple's internal "Atlas" study assesses smart glasses, signaling potential interest in wearable tech to compete with Meta.
Lyft Inks Robotaxi Deals: Lyft collaborates with Mobileye and May Mobility to introduce robotaxis, advancing its position in autonomous ride-hailing.
Microsoft Tests Xbox AI Chatbot: Microsoft’s new AI chatbot assists Xbox Insiders with support, adding personalization options for enhanced user experience.
OpenAI Hires Meta’s AR Lead: OpenAI appoints Caitlin Kalinowski to lead robotics and hardware, indicating plans for AI-driven consumer products.
🧰 TOOLBOX
AI-Powered Tools Transforming Workflows, Insights, Meetings, and Media Creation
AutoGPT | Workflow Automation: AutoGPT enables users to deploy autonomous AI agents for complex workflows with minimal oversight. It supports self-hosting or cloud use, offering ready-made agents for efficient automation.
Dovetail | Customer Insights Hub: Dovetail 3.0 uses AI to turn customer feedback into actionable insights, offering features like automated ticket analysis and a consumer recruitment pool for streamlined organizational knowledge-sharing.
MeetGeek | Meeting Assistant: MeetGeek automates meeting notes and enhances productivity by providing role-specific insights, CRM syncing, and secure integrations, making meetings more efficient for teams across various departments.
PodLM | Podcast Creation: PodLM converts content into polished podcasts with AI-driven script generation, customizable voices, and multi-speaker support, simplifying professional podcast creation for all skill levels with easy publishing.
📽️ INTERVIEW
ASI/AGI, Open vs Closed AI, Founding with Sam Altman, and Verifying Humanness!
In case you missed it, I recently interviewed Saturnin Pugnet, the visionary founder of WorldID. In this conversation, we explore his mission to create a future where people can securely verify their human identity in a world increasingly populated by AI. We dive into the groundbreaking infrastructure behind WorldID and what it means for the future of digital identity, privacy, and trust. Check out the full interview below! 👇
🤔 FRIDAY FACTS
How Much Energy Does a ChatGPT Query Use?
A single ChatGPT query consumes approximately 2.9 watt-hours (Wh) of electricity, which is roughly equivalent to the power needed for 10 Google searches.
This may seem small, but with millions of queries processed every day, the collective energy usage becomes substantial. For instance, if ChatGPT processes 10 million queries in a day, that’s roughly equivalent to the average daily electricity consumption of over 3,200 U.S. households.
🗒️ FEEDBACK
Help Us Get Better
What did you think of today's newsletter? |
CONNECT
Stay in the Know
Follow us on X for quick daily updates and bite-sized content.
Subscribe to our YouTube channel for in-depth technical analysis.
Prefer using an RSS feed? Add Forward Future to your feed here.
Thanks for reading today’s newsletter. See you next time!
The Forward Future Team
🧑🚀 🧑🚀 🧑🚀 🧑🚀
