Commentary and a selection of the most important recent news, articles, and papers about AI.
Today’s Brief Commentary
Most of the links today concern AI agents, autonomous processes acting in the background on your behalf, or so you think. The comparison to Skynet from the Terminator movies is pretty clear, though no one wants to mention that. While we are not there yet, there is certainly a line that is crossed from AI giving you advice to actually performing actions without your permission. This needs governance, extensive testing, fail safes, and clear explanations from providers that describe what can happen and why. Contracts might have to be amended to include an “undo” clause for some time period so that agent actions can be reversed.
General News, Articles, and Analyses
Investments in generative AI startups topped $3.9B in Q3 2024 | TechCrunch
https://techcrunch.com/2024/10/20/investments-in-generative-ai-startups-topped-3-9b-in-q3-2024/
Author: Kyle Wiggers
(Sunday, October 20, 2024) “Not everyone is convinced of generative AI’s return on investment. But many investors are, judging by the latest figures from funding tracker PitchBook. In Q3 2024, VCs invested $3.9 billion in generative AI startups across 206 deals, per PitchBook. (That’s not counting OpenAI‘s $6.6 billion round.) And $2.9 billion of that funding went to U.S.-based companies across 127 deals.”
Agentic AI
Waiting for Skynet | Center for Strategic and International Studies
https://www.csis.org/analysis/waiting-skynet
Author: James Andrew Lewis
Commentary:
This article is from 2018, but it is worth noting that people were concerned about agents and autonomous AI long before generative AI. It also allows me to mention SkyNet, which I know some of you were thinking about.(Thursday, January 18, 2018) “Computers were invented to augment human performance. They are powerful tools, but even as processing speeds increase and algorithms grow more sophisticated, these machines still cannot “think.” Eventually this will change. A group of leading scientists and public figures signed an open letter warning of the dangers of this moment. One famous scientist warned that “The development of full artificial intelligence could spell the end of the human race.””
Agentic AI: Decisive, operational AI arrives in business | CIO
https://www.cio.com/article/3496519/agentic-ai-decisive-operational-ai-arrives-in-business.html
Author: Grant Gross
(Friday, August 30, 2024) “Agentic AI, at its core, is designed to automate a specific function within an organization’s myriad business processes, without human intervention. AI agents can, for example, handle customer service issues, such as offering a refund or replacement, autonomously, and they can identify potential threats on an organization’s network and proactively take preventive measures.”
The era of AI agents is just getting started | Axios
https://www.axios.com/2024/10/21/ai-agents-microsoft-salesforce-sierra
Author: Ina Fried
(Monday, October 21, 2024) “Giving autonomy to generative AI tools opens up a range of tantalizing possibilities for increased productivity, but also vastly increases the potential of catastrophic risk.”
Microsoft introduces ‘AI employees’ that can handle client queries | Microsoft | The Guardian
Author: Dan Milmo
(Monday, October 21, 2024) “Microsoft is introducing autonomous artificial intelligence agents, or virtual employees, that can perform tasks such as handling client queries and identifying sales leads, as the tech sector strives to show investors that the AI boom can produce indispensable products.”
AI in Games
The promise (and pitfalls) of NPCs powered by AI in video games | TechSpot
https://www.techspot.com/news/105202-promise-pitfalls-chatgpt-powered-npcs-modern-video-games.html
Author: Cal Jeffrey
(Friday, October 18, 2024) “Large language models and generative AI are topics that most video game developers would rather avoid. As tempting as using these tools is to replace human labor, the negative blowback is far too intense for most companies to handle, and that’s not even considering that AI technology is not quite at the point where it can consistently produce quality content without human assistance. However, such barriers don’t exist for regular folks. People are already experimenting with AI technology in existing games. Modding communities have begun using platforms such as ChatGPT to give voice to NPCs and followers in games like Skyrim and Stardew Valley.”
Generative AI and Models
IBM Introduces Granite 3.0: High Performing AI Models Built for Business – Oct 21, 2024
(Monday, October 21, 2024) “The new Granite 3.0 8B and 2B language models are designed as ‘workhorse’ models for enterprise AI, delivering strong performance for tasks such as Retrieval Augmented Geneneration (RAG), classification, summarization, entity extraction, and tool use. These compact, versatile models are designed to be fine-tuned with enterprise data and seamlessly integrated across diverse business environments or workflows.”
Granite | IBM
“Start building with Granite 3.0, our family of open, performant and trusted AI models, tailored for business and optimized to scale your AI applications.”
Technical Papers, Articles, and Preprints
[2410.15489] Generative AI Agents in Autonomous Machines: A Safety Perspective
https://arxiv.org/abs/2410.15489
Authors: Jabbour, Jason and Reddi, Vijay Janapa
(Sunday, October 20, 2024) “The integration of Generative Artificial Intelligence (AI) into autonomous machines represents a major paradigm shift in how these systems operate and unlocks new solutions to problems once deemed intractable. Although generative AI agents provide unparalleled capabilities, they also have unique safety concerns. These challenges require robust safeguards, especially for autonomous machines that operate in high-stakes environments. This work investigates the evolving safety requirements when generative models are integrated as agents into physical autonomous machines, comparing these to safety considerations in less critical AI applications. We explore the challenges and opportunities to ensure the safe deployment of generative AI-driven autonomous machines. Furthermore, we provide a forward-looking perspective on the future of AI-driven autonomous systems and emphasize the importance of evaluating and communicating safety risks. As an important step towards addressing these concerns, we recommend the development and implementation of comprehensive safety scorecards for the use of generative AI technologies in autonomous machines.”
[2410.16128] SMART: Self-learning Meta-strategy Agent for Reasoning Tasks
https://arxiv.org/abs/2410.16128
Authors: Liu, Rongxing; Shridhar, Kumar; Prajapat, Manish; Xia, Patrick; and Sachan, Mrinmaya
(Monday, October 21, 2024) “Tasks requiring deductive reasoning, especially those involving multiple steps, often demand adaptive strategies such as intermediate generation of rationales or programs, as no single approach is universally optimal. While Language Models (LMs) can enhance their outputs through iterative self-refinement and strategy adjustments, they frequently fail to apply the most effective strategy in their first attempt. This inefficiency raises the question: Can LMs learn to select the optimal strategy in the first attempt, without a need for refinement? To address this challenge, we introduce SMART (Self-learning Meta-strategy Agent for Reasoning Tasks), a novel framework that enables LMs to autonomously learn and select the most effective strategies for various reasoning tasks. We model the strategy selection process as a Markov Decision Process and leverage reinforcement learning-driven continuous self-improvement to allow the model to find the suitable strategy to solve a given task. Unlike traditional self-refinement methods that rely on multiple inference passes or external feedback, SMART allows an LM to internalize the outcomes of its own reasoning processes and adjust its strategy accordingly, aiming for correct solutions on the first attempt. Our experiments across various reasoning datasets and with different model architectures demonstrate that SMART significantly enhances the ability of models to choose optimal strategies without external guidance (+15 points on the GSM8K dataset). By achieving higher accuracy with a single inference pass, SMART not only improves performance but also reduces computational costs for refinement-based strategies, paving the way for more efficient and intelligent reasoning in LMs.”
