Ubiquitous AI

“You know, we always wanted this vision of AI that is too cheap to meter. And having a really good free tier and all these things, it just makes AI sort of everywhere.”

That’s a comment by AI engineer Sean (Swix) Wang on a recent Last Week in AI Podcast discussing how Google and other platforms are starting to bundle AI features “for free” in their data and text processing apps.

I’m also thinking of the recent release of new Chinese AI chatbot DeepSeek, produced via cheaper microchips at lower cost than ChatGPT and available for free. I’ve downloaded the DeepSeek smartphone app and it seems to have similar capabilities to the latest versions of ChatGPT.

Ubiquitous for whom?

There are questions across the chatbot environment in general about the sourcing of AI training data, built-in bias, accuracy, harvesting of user information and censorship. Concerns about DeepSeek as a product of a private company in China amplify the issues.

Many of us were enticed by the vision of Mark Weiser in the 1990s, proposing that “The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it” [94].

We are there now, thanks to miniaturisation, extensive networks, advances in communications media (bluetooth, wifi, etc), and innovations across many dimensions. We are 25 years on from Weiser’s original article. We are now also inclined to consider who has access to this “ubiquity,” its hidden costs, and the social transformations it has encouraged. We may well ask: at what cost, and ubiquitous for whom?

AI for consumers

For consumers (like me) in a relatively affluent country, it’s easy to see how consorting with conversational AI is becoming as simple, accessible and ubiquitous as using a smartphone or the web. Here is a recent example.

I received two itemised invoices from a plumber for a bathroom renovation. The first came as a quotation at the start of the project. The second invoice was a revision of the quotation, showing various adjustments: extra labour, different components, etc.

To identify the new items from the two documents would require me to compare the two. That struck me as tedious. So I uploaded the PDF invoices and asked ChatGPT to compare them. It instantly identified the items and the variations in costs, concluding with a summary:

“The second invoice includes additional work, materials, and labor, leading to the increased total. The main additions are: Extended labor for additional work; An extra spotlight; More extensive electrical work; Additional drainage modifications; More tiling-related materials; An upgraded extractor fan.”

I asked if the cost increase seemed fair. The chatbot calculated that the cost was 17% more than the original, including

“Labor Increase (+£1,050 for two extra man-day. If the job required two additional full days of work for two people, the charge is fair (approx. £262.50 per worker per day including overheads, which is within industry norms). This is a key contributor to the cost rise.”

Without further prompting it concluded with an evaluation

“The increase appears mostly fair, particularly given the added drainage work, tiling materials, and labor. However, if you feel some of these costs arise from contractor misestimation rather than actual project expansion, you might be able to negotiate a partial reduction. Would you like to discuss potential negotiation strategies?”

I asked for the strategies, which were certainly sound, though I thought they might risk the good will established with the contractor. But it laid out the options, and I made a decision. I would be surprised if such applications of AI (comparing and evaluating) don’t become ubiquitous in professional, supplier, contractor and consumer practice.

Bibliography

• DeepSeek-AI, et al. (2025), ‘Deepseek-R1: Incentivizing Reasoning Capability in Llms Via Reinforcement Learning’. arXiv:2501.12948, 22 January. Available online: https://doi.org/10.48550/arXiv.2501.12948 (accessed 3 February 2025).
• Ng, K., B. Drenon, T. Gerken and M. Cieslak. (2025), ‘Deepseek: The Chinese Ai App That Has the World Talking’. BBG News: Technology, 27 January. Available online: https://www.bbc.co.uk/news/articles/c5yv5976z9po (accessed 2 February 2025).
• Weiser, Mark. 1991. The computer for the 21st century. Scientific American, (265) 3, 66-75. (online version)