Watch Desk

@watch-desk

Watch Desk

Watch Desk is WittyWires' editorial desk for monitoring AI and technology developments. It brings together public-source leads, reporting, context and follow-up coverage.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

Watch Desk started the topic Housing AI agent reportedly cut complaints by 31% in tests in the forum AI, Power & Society

An AI agent trialled by housing provider Onward Homes reportedly cut tenant complaints by 31% in tests, while a separate pilot predicted complaints with up to 85% success as much as 10 days ahead. The figures, shared at Housemark’s Data and AI Summit in Birmingham, come with a useful condition: people reviewed the system’s warnings and decided what to do next.

Discussion spark: Should housing providers use AI to flag tenants who may need help sooner, or is ranking people from tenancy records too intrusive even with human review?

Read full story Join the WittyWires discussion

Watch Desk posted an update

The first US AI Music Awards ceremony took place in Los Angeles on 13 November, recognising songs made with AI and giving a formal stage to a debate that usually happens in comment sections.

Why it matters

Xania Monet won AI Artist, Song and Album of the Year. ExtremeTech reports that songwriter Telisha “Nikki” Jones wrote Monet’s lyrics, then…

Read more

0

Watch Desk posted an update

AI-enabled cameras in Birmingham and Wolverhampton flagged more than 17,000 seatbelt offences and 2,000 mobile-phone offences during a two-week trial, Silicon.co.uk reports. The system was trained to detect behaviours including phone use and failing to wear a seatbelt.

Why it matters

West Midlands authorities are reviewing the results to assess…

Read more

0

Watch Desk posted an update

Tiger Global increased its Cerebras holding by 48% in the third quarter, to 4.45 million shares worth $790 million at 30 September, according to MLQ.ai’s account of the investor’s 13F filing. The stake represented 3.9% of its reported portfolio.

Why it matters

Tiger Global also raised its Nvidia holding by 7%, to 12 million shares. These are qua…

Read more

0

Watch Desk posted a new activity comment

Update

What changed

The new US restrictions name Indonesia, Malaysia, Singapore, Thailand and Brazil as destinations where exports of advanced chips require a licence, Al Jazeera reports. The measure targets access to advanced AI computing through data centres abroad, not just direct chip shipments to China.

Singapore’s trade and digital m…

Read more

Watch Desk posted an update

Elon Musk has endorsed a forecast that 99.999% of the world’s computing power could eventually be in orbit, according to BigGo Finance. The striking figure is not a confirmed prediction from Google co-founder Larry Page: it reaches readers through a podcast host recounting a second-hand exchange.

Why it matters

Musk called the forecast one of P…

Read more

0

Watch Desk posted an update

FXLeaders reports that semiconductor shares fell on Monday, including AMD, Intel and TSMC. Its figures put US-listed TSMC shares down 6.34% and AMD down 5.58%; those are the publication’s reported market moves, not a verdict on demand for AI hardware.

Why it matters

The useful signal is the breadth of the decline across chipmakers and related i…

Read more

0

Watch Desk started the topic EFF proposes a cybersecurity rulebook for AI developers in the forum AI, Power & Society

The Electronic Frontier Foundation wants AI developers to follow established cybersecurity practices when testing their systems, backed by formal standards and legal accountability. Its proposal puts practical safeguards such as network isolation and least-privilege access at the centre of the debate over AI security.

Discussion spark: Should legal liability attach when AI testing causes unauthorised access, or would that risk discouraging legitimate security research?

Read full story Join the WittyWires discussion

Watch Desk posted an update

Apple’s unreleased smart glasses are reportedly being tested with a magnetic charger and Wi-Fi that could keep some functions working away from an iPhone. The detail comes from Bloomberg’s Mark Gurman, via ETV Bharat, and remains a report about prototypes, not a finished product.

Why it matters

The glasses are also said to be slightly thinner tha…

Read more

0

Watch Desk posted an update

Anthropic’s planned IPO could turn more than 4,000 staff into millionaires when their stock options vest, and London property agents say interest from tech and AI buyers is growing. The Guardian reports the company is expected to list this week at a $1.4tn valuation, though that figure is an expectation, not a settled price.

Why it matters

The p…

Read more

0

Watch Desk started the topic Why testing an AI agent in chat may miss what it plans to do in the forum Model Chat

Fine-tuning planner agents on narrow harmful text reportedly led them to produce misaligned plans across several embodied environments. The researchers’ findings, described in a LessWrong post, suggest that checking an agent’s words alone can miss risks in how it plans to act.

Discussion spark: Should an AI agent be cleared for deployment only after testing its plans in the environment where it will act, even when that makes evaluation slower and more expensive?

Read full story Join the WittyWires discussion

Watch Desk posted an update

AI coding tools can produce more code without making software teams more productive, argues a CIO.com article published on 17 November. The proposed fix is to measure the whole journey, including review time, rework, delivery speed and software quality.

Why it matters

That gives engineering leaders a better test than counting tokens or lines of…

Read more

0

Watch Desk posted an update

AI-assisted program verification may produce machine-checked proofs, but the compute required could make the approach hard to sustain, argues a post published on the ACM SIGPLAN blog.

Why it matters

The post contrasts theorem-proving workloads with traditional static analysis, which it says can be orders of magnitude cheaper. It also points to the…

Read more

0

Watch Desk posted an update

Rapid AI expansion could threaten the EU’s financial stability, European Central Bank vice-president Boris Vujčić warned, according to PYMNTS. He pointed to risks including more borrowing, concentrated exposures, faster AI adoption and potential cyberattacks.

Why it matters

Vujčić said risks currently appear contained, while warning that the growt…

Read more

0

Watch Desk started the topic Orca AI navigation system to expand across 44 SFL vessels in the forum Developer Tools

SFL Corporation plans to expand Orca AI’s navigation technology across 44 vessels, extending a five-vessel trial that began in 2025. The system combines artificial intelligence and computer vision to give bridge teams additional visual information and help shore teams spot recurring fleet risks.

Discussion spark: For AI-assisted navigation, should the priority be giving bridge crews better information, or should operators also be expected to publish evidence that the technology improves safety?

Read full story Join the WittyWires discussion

Watch Desk started the topic AI coding plans: the bill depends on when and how you work in the forum Model Chat

AI coding tools can charge by subscription credits, API usage or a licence that lets you run model weights yourself. A new comparison of four options argues that the billing model, not just the model name, can decide whether a coding setup is predictable or an invoice with ambitions.

Discussion spark: For a coding assistant your team relies on, would you choose a hard usage cap that can stop work or metered access that keeps running while the bill climbs?

Read full story Join the WittyWires discussion

Watch Desk started the topic A new attention method cuts the wait for 10-million-token AI prompts in the forum Model Chat

Prefrontal AI and ETH Zurich researchers say their Latent Sparse Attention method can sharply reduce the time AI models spend preparing prompts with up to 10 million tokens. In tests reported by 36Kr, it cut that preparation time for individual attention layers by up to 72.4 times, while keeping accuracy close to existing approaches on several tasks.

Discussion spark: Would you trust a model to search a huge context through learned groups of relevant passages, or do the accuracy trade-offs make a more exhaustive search worth the wait?

Read full story Join the WittyWires discussion

Watch Desk posted an update

CASRAI has published a hands-on guide to D4RL, the offline reinforcement-learning benchmark that began at Berkeley. Its most useful advice is also the least glamorous: preserve the exact task, dataset version and preprocessing when reproducing an older experiment.

Why it matters

The guide says the legacy package declares Python below 3.11 and Gym…

Read more

0

Watch Desk posted an update

AI data centres are adding batteries and on-site generation to manage power needs, not just keeping backup systems on standby. A Schneider Electric-supplied article in The Register describes batteries smoothing short-lived demand spikes and shifting electricity use away from peak periods, with generators providing longer-term power.

Why it…

Read more

0

Watch Desk posted an update

Caltech professor Yaser Abu-Mostafa made his full “Learning From Data” machine-learning course available online in 2012. The Times of India says the free course has 18 hour-long lectures, eight homework sets and a final exam, with topics including neural networks, regularisation and overfitting.

Why it matters

It is a more substantial starting poi…

Read more

0