Commentary and a selection of the most important recent news, articles, and papers about AI.
Today’s Brief Commentary
Two of the links today are related to an IBM survey of software developers regarding the creation of generative AI apps. TLDR: it’s not easy, and the tools aren’t great.
Developers tend to build the tools they need, and we should expect some very good open-source ones to be extended or created in 2025. Of course, as the core GenAI systems evolve, the tools will need to keep pace. These will go beyond AI code generation support facilities, of which we now seem to have dozens. I’m primarily talking about tools for AI development here, not tools that use AI. The areas do overlap.
I want to highlight another category of links today: federated learning. As the examples from 2017 and 2019 show, this is not a new idea. Nevertheless, generative AI practitioners are now pulling in techniques from machine learning that were commonly used more than five years ago. We are fine as long as they don’t claim they have newly invented some older technique. Conversely, not using a pre-GenAI idea because of a lack of awareness is just sloppy professional research.
Always ask: “Has anyone ever done something like this in another context I can adopt to solve my problem?”. Great developers do this all the time. New or middling developers can improve by explicitly putting their curiosity on overdrive.
Contents
- AI and the Media
- Coding and Software Engineering
- Games
- Generative AI and Models
- Predictions
- Responsible AI
- Research and Technical
- Related Articles and Papers
AI and the Media
An update on Generative AI (Gen AI) at the BBC
https://www.bbc.com/mediacentre/2025/articles/update-generative-ai-at-the-bbc
Author: Rhodri Talfan Davies
Date: Thursday, January 16, 2014
Commentary: This is an interesting update, but surely there were some things that didn’t work out as planned.
Excerpt: What have we learned so far? The Gen AI tools are impressive, the accuracy levels are improving and the pilots suggests we could unlock significantly more value for audiences. In the majority of cases, it’s also clear AI can assist us to do things more quickly but it does not remove the need for human oversight given the risk of error.
Coding and Software Engineering
Survey: Generative AI Makes Tasks Simple, But Developing That AI is Anything But | IBM
Author: Ritika Gunnar
Date: Wednesday, January 8, 2025
Excerpt: Now, a new survey sponsored by IBM and conducted by Morning Consult explores that complexity – and unveils the challenges developers are facing when it comes to skills variance, vast and complicated toolsets, and ensuring accurate and trusted results from these systems. Our survey interviewed more than 1,000 enterprise AI developers in the U.S. who are building generative AI applications for the enterprise. Survey participants span a range of roles, including application developer, software engineer, and data scientist.
Developers are at their wits end trying to build generative AI applications – skills gaps, complexity, and ‘tool sprawl’ are creating major hurdles | ITPro
Author: George Fitzmaurice
Date: Monday, January 13, 2025
Commentary: This article is based on the IBM survey above, but provides some third-party insight. It is imperative and non-debatable that any kind of engineer must maintain, update, and broaden their skills. However, don’t blame them if their provided tools are substandard and their management does not understand the problem.
Excerpt: AI coding tools could help, though experts told ITPro that businesses should keep a focus on core software engineering skills
Games
Microsoft Files AI Patent For Player-Driven Video Game Storylines | WinBuzzer
Author: Markus Kasanmascheff
Date: Tuesday, January 14, 2025
Commentary: This should have an interesting response from the game dev community.
Excerpt: Microsoft has filed a patent showcasing how generative AI can dynamically create game narratives, integrating real-time player feedback and engagement metrics.
Generative AI and Models
Federated learning: The killer use case for generative AI | InfoWorld
https://www.infoworld.com/article/3804403/federated-learning-the-killer-use-case-for-genai.html
Author: David Linthicum
Date: Friday, January 17, 2025
Excerpt: Federated learning is not new, so I’m often surprised by how few people know about it. I’ve included links below to 2017 and 2019 articles that describe how it works. Many vendors have more modern material on the websites.
Predictions
What’s next for AI in 2025 | MIT Technology Review
https://www.technologyreview.com/2025/01/08/1109188/whats-next-for-ai-in-2025/
Authors: James O’Donnell; Will Douglas Heaven; and Melissa Heikkilä
Date: Wednesday, January 8, 2025
Excerpt: So what’s coming in 2025? We’re going to ignore the obvious here: You can bet that agents and smaller, more efficient, language models will continue to shape the industry. Instead, here are five alternative picks from our AI team.
Responsible AI
Explore the business case for responsible AI in new IDC whitepaper | Microsoft Azure Blog
Author: Sarah Bird
Date: Monday, January 6, 2025
Excerpt: This whitepaper, based on IDC’s Worldwide Responsible AI Survey sponsored by Microsoft, offers guidance to business and technology leaders on how to systematically build trustworthy AI. In today’s rapidly evolving technological landscape, AI has emerged as a transformative force, reshaping industries and redefining the way businesses operate. Generative AI usage jumped from 55% in 2023 to 75% in 2024; the potential for AI to drive innovation and enhance operational efficiency is undeniable. However, with great power comes great responsibility. The deployment of AI technologies also brings with it significant risks and challenges that must be addressed to ensure responsible use.
Research and Technical
[2501.05435] Neuro-Symbolic AI in 2024: A Systematic Review
https://arxiv.org/abs/2501.05435
Authors: Colelough, Brandon C. and Regli, William
Date: Thursday, January 9, 2025
Excerpt: Background: The field of Artificial Intelligence has undergone cyclical periods of growth and decline, known as AI summers and winters. Currently, we are in the third AI summer, characterized by significant advancements and commercialization, particularly in the integration of Symbolic AI and Sub-Symbolic AI, leading to the emergence of Neuro-Symbolic AI. Methods: The review followed the PRISMA methodology, utilizing databases such as IEEE Explore, Google Scholar, arXiv, ACM, and SpringerLink. The inclusion criteria targeted peer-reviewed papers published between 2020 and 2024. Papers were screened for relevance to Neuro-Symbolic AI, with further inclusion based on the availability of associated codebases to ensure reproducibility. Results: From an initial pool of 1,428 papers, 167 met the inclusion criteria and were analyzed in detail. The majority of research efforts are concentrated in the areas of learning and inference (63%), logic and reasoning (35%), and knowledge representation (44%). Explainability and trustworthiness are less represented (28%), with Meta-Cognition being the least explored area (5%). The review identifies significant interdisciplinary opportunities, particularly in integrating explainability and trustworthiness with other research areas. Conclusion: Neuro-Symbolic AI research has seen rapid growth since 2020, with concentrated efforts in learning and inference. Significant gaps remain in explainability, trustworthiness, and Meta-Cognition. Addressing these gaps through interdisciplinary research will be crucial for advancing the field towards more intelligent, reliable, and context-aware AI systems.
Related Articles and Papers
Federated Learning: Collaborative Machine Learning without Centralized Training
Authors: Brendan McMahan and Daniel Ramage
Date: Thursday, April 6, 2017
Commentary: This is an older reference and uses mobile phones as the example remote devices.
Excerpt: It works like this: your device downloads the current model, improves it by learning from data on your phone, and then summarizes the changes as a small focused update. Only this update to the model is sent to the cloud, using encrypted communication, where it is immediately averaged with other user updates to improve the shared model. All the training data remains on your device, and no individual updates are stored in the cloud.
[1908.07873] Federated Learning: Challenges, Methods, and Future Directions
https://arxiv.org/abs/1908.07873
Authors: Li, Tian; Sahu, Anit Kumar; Talwalkar, Ameet; and Smith, Virginia
Date: Wednesday, August 21, 2019
Excerpt: Federated learning involves training statistical models over remote devices or siloed data centers, such as mobile phones or hospitals, while keeping data localized. Training in heterogeneous and potentially massive networks introduces novel challenges that require a fundamental departure from standard approaches for large-scale machine learning, distributed optimization, and privacy-preserving data analysis. In this article, we discuss the unique characteristics and challenges of federated learning, provide a broad overview of current approaches, and outline several directions of future work that are relevant to a wide range of research communities.