Hardware is hard to do
Montfort's AI 3-60 Newsletter #11, 23 May 2025
Introduction
This week, LLMs form communities, OpenAI buys IO in a bet on hardware and workers are excited but overwhelmed by Generative AI. We also learn what dot-ai can learn from dot-com and we receive a deluge of announcements from Google I/O.
In this week’s issue:
Study: LLMs form communities and that may be a problem
Study: Employees excited by but overwhelmed by LLM progress
Deals: OpenAI buys IO
Generative AI 3 Minute Lesson: what dot-ai can learn from dot-com
Generative AI 60 minute Lesson: Google I/O Keynote and everything announced
🔎 The Week’s Big AI News at a Glance
🔗 LLM Agents and Social Conventions – A paper submitted to a peer reviewed journal suggests that LLM Agents may spontaneously develop social conventions if left to their own devices, forming the background of what we might understand as a society. In particular, minorities of motivated agents were able to exert their influence over the majority to drive social change. 🔗 The Independent has more on the experiment. We think: 🧠 This study has implications for how we think about alignment of LLMs, and particularly their more autonomous agent variants. Clearly if LLMs can develop organising principles and social norms without explicit programming, this form of emergent behaviour makes alignment monitoring more difficult. It also means that as competing intelligences develop, there is a need to safeguard models from being malignly influenced by other intelligences, not just humans.
🔗 Workers are excited but overwhelmed by AI – A recent Henley Business School survey reports that 56% of UK workers are optimistic of the benefits AI may bring, but 61% find it difficult to remain apace with the rate of technology change. We think: 🧠 Those of you a little overwhelmed – you are not alone. Part of the reason we set up this newsletter was that we recognised the immense potential of AI for our profession, but also the rate of acceleration in the development of this technology. If we can be of further help to you in understanding Generative AI and its implications, please do get in touch via the details at the end of this newsletter.
🔗 Jonny Ive joins OpenAI – In a very unusual deal, OpenAI has acquired IO, the startup that Sam Altman and Jonny Ive have been working on for a couple of years, with the deal valuing the company at $6.5bn. Jonny Ive’s firm Love From will also now assume responsibility for all design work across OpenAI, with IO’s founding purpose to create the next wave of devices that are AI-first. We think: 🧠 The announcement may be timed to coincide with Google’s I/O Conference (see this week’s 60 minute lesson) at which it directly addresses hardware and software developers. However, the valuation is staggering given that there are no products to speak of yet. Yet, as Generative AI becomes more multimodal, and can understand context, mood and environment, the challenges of designing interfaces (software and hardware) become more challenging. Marshalling coherence across modes and designing for trust and transparency operates at the intersection of industrial design, ethics and human-computer interaction. It will have to be something pretty special to unseat the mighty smartphone as the primary portal to information, but with a team comprising many designers of the iPod, iPhone and MacBook Pro, they’ve a better chance than most.
⏲️ Spend 3 Minutes With… (Quick Wins)
🔗 What dot-ai can learn from dot-com – This week, a slightly different tack on the business transformation challenges and opportunities posed by Generative AI. For those of us old enough to recall at least some of the dot-com era, there’s much to learn in terms of what it means to transition to the world of AI. Plenty to learn from in this thought leadership piece from an AI Product Manager at Uber.
🕙 Spend an Hour With (Deeper Dives)
🔗 Everything from Google I/O – Google I/O, the conference where the firm addresses software and hardware engineers took place this week. The top line is that Google is putting more AI into search, email and the browser and hardware in the form of a new pair of glasses. The link takes you to a blow by blow from The Verge, but the full keynote is there if you want to watch it. Also worth it for the appearance by Google Co-Founder Sergey Brin who thinks we live in not just one but 🔗 a stack of simulations. There’s 🔗 a ten minute highlight reel of all the company’s announcements if you prefer.
🚀 Final Thought
Our Research, Data, and Analytics team helps companies across sectors understand how to protect and enhance their reputations using advanced analytics and machine learning. We offer workshops for communications teams looking to get ahead, as well as tools and techniques to navigate AI’s impact on reputation. If you’re interested, get in touch at digital@montfort.london.
Patrick, Owen and Ben at Montfort