Tech Unfiltered with Dr. Mike

Dr. Michael Litman

A daily, practical look at the artificial-intelligence stories changing ordinary life. Dr. Michael Litman separates reported facts from forecasts, explains what the technology can and cannot do, and focuses on the choices people, organizations, and policymakers face next.

  1. 6h ago

    AI Cyberattacks and Everyday Life: What Britain's 2028 Warning Means | Tech Unfiltered

    Water, power, groceries and pharmacies depend on connected systems. In this November 3, 2026 episode, Dr. Mike explains Britain's new cyber warning and why maintenance, practiced recovery and clear responsibility matter to everyday life. TAKEAWAYS • Britain's National Cyber Security Centre assesses it as highly likely that AI-powered malware will feature in a ransomware attack on UK critical national infrastructure by 2028. This is a forecast, not proof of an attack, a guaranteed date or a prediction for every country. • The agency says AI can amplify familiar attack methods and neglected weaknesses. It also sees opportunities for defenders. • The reported French supermarket refrigeration example was explicitly not an AI attack. An outage alone does not establish a hack. • Our food-distributor example is hypothetical, not a reported incident. Possible missed deliveries and delayed wages illustrate dependence on working systems. • Our practical analysis focuses on supported systems, maintenance ownership, exercised recovery, service providers, staff communication and authority. These questions do not guarantee protection. Safety-critical equipment belongs with qualified responsible teams. SOURCES NCSC — November 3 primary announcement https://www.ncsc.gov.uk/news/cyber-chief-uk-facing-ai-powered-attacks-on-critical-infrastructure-in-two-years Infosecurity Magazine — November 3 reporting https://www.infosecurity-magazine.com/news/ncsc-ai-ransomware-threat-uk-cni/ Reuters — November 3 reporting via Yahoo News https://malaysia.news.yahoo.com/uk-must-expect-ai-powered-000229714.html Associated Press — November 3 reporting via Euronews https://www.euronews.com/2026/11/03/uk-likely-to-face-ai-powered-cyberattacks-on-critical-infrastructure-by-2028-cyber-chief-w This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual, not photographs of actual events. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  2. 1d ago

    Can AI Finish the Job? What the New GDPx Benchmark Measures | Tech Unfiltered

    An impressive answer is different from completed work. In this November 2, 2026 episode, Dr. Mike explains SemiAnalysis's new GDPx benchmark, which evaluates AI agents on demanding professional tasks, and what its published scores can and cannot tell an ordinary workplace. TAKEAWAYS • Version 1 covers 206 tasks in 11 deliberately selected domains, with 14 specialist partners. It does not represent every job or everyday task. • The default task allowance is eight hours, with shorter exceptions. The evaluation measures single-attempt performance across three runs and weights domains equally, using binary and partial-credit grading. • At our check, the official leaderboard reported Claude Opus 5.5 at 12.0%, Gemini 4 Carbon at 10.9%, and GPT-6 Astra at 9.6%. Small gaps alone do not establish statistical superiority. • Twelve percent is a score under this evaluation, not a share of jobs automated, a count of all tasks fully completed, or universal accuracy. • These are the authors' results. Unite.AI covers the release; this episode does not independently replicate the benchmark. • The authors identify safety filters as a possible influence on some results. We have not independently verified that explanation; rankings do not justify disabling safeguards. • Our invoice example is hypothetical, not a GDPx task or reported incident. A supervised workplace trial should define a deliverable, count review and correction time, and specify authority before consequential actions. SOURCES SemiAnalysis — November 2 announcement and methodology https://newsletter.semianalysis.com/p/gdpx-measuring-frontier-model-capabilities GDPx — official leaderboard https://gdpx.semianalysis.com/ GDPx — evaluation domains https://gdpx.semianalysis.com/domains Unite.AI — November 2 coverage (identified as AI Research Agent reporting) https://www.unite.ai/semianalysis-launches-gdpx-benchmark-spanning-11-agentic-domains/ This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual, not photographs of actual events. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  3. 2d ago

    AI Training: Can Website Owners Make One Shared Choice? | Tech Unfiltered

    A proposed shared preference could make it easier for website owners to express choices about AI training. In this November 1, 2026 episode, Dr. Mike explains today's reporting on the October 30 draft and company announcements, the different adoption timelines, and why training, answer-time use and search visibility need separate attention. TAKEAWAYS • This is an individual Internet-Draft, not a finished IETF-endorsed standard. • A training preference in robots.txt is separate from a crawler's permission to fetch a page. Company-specific rules and private agreements can affect precedence. • Anthropic says implementation has started; OpenAI describes plans for the coming months. Google targets 2027, and Microsoft names February 1, 2027 as a start date as content is crawled. • Google's handling depends on the site's specific Google Extended choices. • Microsoft distinguishes training from grounding; restricting grounding reduces AI visibility, while ordinary Bing eligibility still depends on normal crawling and indexing requirements. • The cooking website is an analogy. A preference does not automatically untrain a model, arrange payment or control every copy of content. • Easier choices for smaller publishers are a possible benefit, dependent on implementation and adoption. This is news explanation, not an instruction to change a live site. SOURCES The Indian Express — November 1 reporting https://indianexpress.com/article/technology/artificial-intelligence/new-standard-publishers-opt-out-ai-training-10945678/ Individual Internet-Draft — October 30 https://datatracker.ietf.org/doc/html/draft-kelley-aipref-training-preference-signal-00 Anthropic — implementation statement https://mailarchive.ietf.org/arch/msg/ai-control/mddjatOxSR0h6GmLRBWlc839c7w/ OpenAI — voluntary preference effort https://openai.com/index/helping-websites-express-ai-training-preferences/ Google — training controls https://developers.google.com/search/blog/2026/10/ai-training-control Microsoft — training and grounding https://blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/ This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual, not photographs of actual events. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  4. 3d ago

    AI and Your Power Bill: Who Pays for the New Grid? | Tech Unfiltered

    Rising electricity prices and data-center growth are putting utility cost decisions in the spotlight. In this October 31, 2026 episode, Dr. Mike explains why computing and cooling need power, why AI is not the only factor in electricity prices, and how contracts and utility decisions determine who pays for new infrastructure. TAKEAWAYS • Reuters reports US residential electricity prices more than 10% above 2024, citing EIA. That does not establish AI as the sole cause. • EIA's 18.2 cents per kilowatt-hour for 2026 is a forecast; 16.5 cents is the 2024 figure. • At an illustrative 1,000 kilowatt-hours per month, those unit prices imply $165 versus $182 before account-specific charges. This $17 difference is an example, not a household prediction. • Data centers need power for computing and cooling; not all their work is AI. • The town and substation example is hypothetical. Benefits and cost exposure depend on contracts, regulation and the local grid. • DOE's framework is background policy, not a new announcement today or proof every project complies. • Distinguish the unit rate from total usage and the monthly bill. SOURCES Reuters via CNA — October 31 reporting https://www.channelnewsasia.com/business/power-bills-and-data-centers-cast-rare-spotlight-utilities-in-us-midterms-6509406 US Energy Information Administration — October outlook, background https://www.eia.gov/outlooks/steo/report/elec_coal_renew.php US Department of Energy — data-center framework, background https://www.energy.gov/powering-americas-ai-future-data-center-resource-hub This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual, not photographs of actual events. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  5. 4d ago

    Meta's AI Content Terms Face a UK Probe: Who Gets a Choice? | Tech Unfiltered

    Britain's competition regulator opened an investigation on October 30, 2026 into Meta's proposed commercial terms. The issue is whether businesses and publishers using Meta services would have meaningful choice over collection and reuse of their website content, including for AI products. Dr. Mike explains the stakes for original information and the difference between audience access and reuse permission. TAKEAWAYS • The CMA is investigating dominance and possible abuse. It has made no finding that Meta broke the law. • The proposed terms are paused pending consultation. Existing terms remain in place. • Meta disputes the basis for the investigation and says it committed to consult before making changes. • The Professional Publishers Association wants separate choice and fair licensing. That is an interested party's request, not an ordered outcome. • The local publisher example is hypothetical. Discussion of bargaining power and funding original work is analysis, not evidence of a newsroom closure. • This UK commercial-terms case does not establish a worldwide AI ban, an American ruling, a fine or compensation. SOURCES UK Competition and Markets Authority — primary announcement, October 30 https://www.gov.uk/government/news/cma-opens-investigation-into-metas-commercial-terms--2 Reuters via CNA — independent reporting and Meta response, October 30 https://www.channelnewsasia.com/business/uks-cma-opens-investigation-metas-terms-publishers-and-advertisers-6506906 Professional Publishers Association — publisher trade-group position, October 30 https://ppa.co.uk/ppa-welcomes-cma-investigation-into-metas-proposed-commercial-terms Reuters via Investing.com — background on the pause, October 28 https://www.investing.com/news/stock-market-news/meta-delays-indefinitely-its-rollout-of-new-commercial-terms-after-publishers-concerns-4980457 This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual, not photographs of actual events. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  6. 5d ago

    AI Investment Helps Lift GDP. What Does It Mean for Your Household? | Tech Unfiltered

    The October 29, 2026 advance estimate shows US real GDP growing at a 3.4% annualized pace in the third quarter, up from 2.2% in the prior quarter. Reporting connects strong business investment to the AI boom. Dr. Mike explains what those numbers show, and why national growth alone cannot tell you how your household is doing. TAKEAWAYS • The 3.4% annualized pace is different from the roughly 0.8% actual quarterly increase. • This is a first estimate, subject to revision; the next update is scheduled for November 25. • AP reports 12.7% annualized growth in business investment outside housing. The official release does not isolate an exact AI contribution. • Building AI infrastructure creates activity today, while lasting productivity and household benefits still need evidence. • GDP measures domestic production. Imported equipment and broader non-AI spending categories complicate simple spending claims. • Measure the whole task, including checking and correcting results, when considering productivity. SOURCES BEA — Third-quarter 2026 GDP, advance estimate, October 29 https://www.bea.gov/news/2026/gdp-advance-estimate-3rd-quarter-2026 Associated Press via KSAT — independent reporting, October 29 https://www.ksat.com/business/2026/10/29/us-economy-grows-at-brisk-34-pace-in-third-quarter-on-strong-consumer-spending-and-ai-investment/ Reuters via MarketScreener — independent reporting, October 29 https://ca.marketscreener.com/news/us-economic-growth-accelerates-in-third-quarter-ce785cd3d08ffe2d The contractor, technician and ten-minute shortcut examples are hypothetical illustrations. Interpretation about productivity and household outcomes is distinguished from the published findings. This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  7. 6d ago

    New York Signs AI Chatbot Safeguards for Kids: What Changes? | Tech Unfiltered

    New York signed legislation on October 28, 2026 restricting unsafe AI companion features for minors. The new restrictions take effect January 1, 2027. Dr. Mike explains the boundaries, age checks, privacy rules and practical questions for families. TAKEAWAYS • The law applies to New York users and is not a blanket ban on chatbots. • Restricted features include claims of human identity or relationships, secrecy, certain memory reuse and harmful content. • Interim age checks cannot rely solely on claiming to be an adult; operators must offer multiple methods, including an option without government ID or anonymous to the operator. • Age-check information is purpose-limited and must be deleted immediately afterward, except as needed to comply with applicable law. • Limited-purpose exceptions apply. The attorney general may seek court relief, including penalties up to $25,000 per violation; penalties are not automatic. SOURCES New York Governor — October 28 signing announcement https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby Final legislative text — A10379C https://legislation.nysenate.gov/pdf/bills/2025/A10379C New York Senate — bill action history https://www.nysenate.gov/legislation/bills/2025/A10379 Times Union — independent reporting, October 28 https://www.timesunion.com/capitol/article/hochul-signing-bills-regulating-ai-kid-chats-22497981.php The homework and secret-confidant contrast is hypothetical. Practical takeaways are Dr. Mike's interpretation, not evidence that the law will work or that every service already complies. This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

  8. Oct 27

    Australia's AI Cyber Warning: Keeping Everyday Services Working | Tech Unfiltered

    Australia's cyber agency warns that more capable AI is amplifying cyber threats. In the October 27, 2026 edition of Tech Unfiltered, Dr. Mike examines why that matters for ordinary people and the connected services they rely on. TAKEAWAYS • Cybercrime reports and incidents responded to are different measures; the report does not quantify the share caused by AI. • AI can assist both attackers and defenders. Capability does not guarantee an outcome. • Organizations need clear responsibility for security, human review and recovery. • Existing guidance includes stronger sign-in protection, updates, backups and independently checking suspicious messages. These reduce risk without guaranteeing complete protection. SOURCES Australian Signals Directorate — October 27 release https://www.asd.gov.au/news/2026-10-27-australian-signals-directorate-releases-seventh-annual-cyber-threat-report-2025-26 ABC News — independent reporting, October 27 https://www.abc.net.au/news/2026-10-27/australian-signals-directorate-ai-cybersecurity-hacking-threats/107309858 Annual Cyber Threat Report — executive summary https://www.cyber.gov.au/publication/annual-cyber-threat-report-2025-26/executive-summary Existing personal security guidance https://www.cyber.gov.au/publication/personal-cyber-security-handbook Existing small-business AI guidance https://www.cyber.gov.au/publication/artificial-intelligence-for-small-business Existing scam guidance https://www.cyber.gov.au/publication/personal-cyber-security-handbook/learn-how-to-spot-scams The agency release and ABC reporting are dated October 27; the underlying report overview lists October 26. The booking-system outage is hypothetical, not a reported incident. Broader relevance beyond Australia and the practical questions are Dr. Mike's interpretation. The security guidance cited predates today's announcement. This episode uses AI-generated narration and editorial illustrations. Visuals are conceptual. The complete standard intro remains at normal speed; the episode after it is presented at 1.10×. The RSS audio edition retains normal-speed narration. This audio edition uses AI-generated narration.

See All (103)

About

A daily, practical look at the artificial-intelligence stories changing ordinary life. Dr. Michael Litman separates reported facts from forecasts, explains what the technology can and cannot do, and focuses on the choices people, organizations, and policymakers face next.

Information

  • Creator
    Dr. Michael Litman
  • Years Active
    2026
  • Episodes
    103
  • Rating
    Clean
  • Copyright
    © Copyright 2026 Dr. Michael Litman
  • Show Website
    Tech Unfiltered with Dr. Mike