![[The AI Show Episode 144]: ChatGPT’s New Memory, Shopify CEO’s Leaked “AI First” Memo, Google Cloud Next Releases, o3 and o4-mini Coming Soon & Llama 4’s Rocky Launch](https://www.marketingaiinstitute.com/hubfs/ep%20144%20cover.png)
Weekly Review 25 April 2025
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):How AI can help with taxes and financial planning: https://www.bigdatawire.com/2025/04/18/the-transformative-role-of-ai-in-financial-planning-and-tax-preparation/Actors who sold their likenesses for AI avatars are regretting doing so: https://arstechnica.com/ai/2025/04/regrets-actors-who-sold-ai-avatars-stuck-in-black-mirror-esque-dystopia/AI is widely used in the New Zealand public sector, mostly because it helps them with their work: https://www.rnz.co.nz/news/political/558081/ai-widely-used-in-public-sector-survey-findsOne-bit neural network weights, to reduce the memory footprints of large AI: https://arstechnica.com/ai/2025/04/microsoft-researchers-create-super%e2%80%91efficient-ai-that-uses-up-to-96-less-energy/The role of Chief AI Officer is expected to become more prominent in coming years: https://www.informationweek.com/machine-learning-ai/how-will-the-role-of-chief-ai-officer-evolve-in-2025-How AI is affecting the creative professions: https://www.nzherald.co.nz/nz/how-ais-latest-advancements-could-reshape-creative-professions-the-front-page/NVIPGTLUR5GVFDBUN4FWJW23EE/AI agent hallucinates new policy, upsets users to the point of cancelling subscriptions: https://arstechnica.com/ai/2025/04/cursor-ai-support-bot-invents-fake-policy-and-triggers-user-uproar/AI models get more advanced, but they just hallucinate more: https://techcrunch.com/2025/04/18/openais-new-reasoning-ai-models-hallucinate-more/Wikipedia got so hammered by AI training bots scraping the site, they have released a data set optimised for training AI to Kaggle: https://www.theverge.com/news/650467/wikipedia-kaggle-partnership-ai-dataset-machine-learningGoogle is using AI to block scammers' ads: https://arstechnica.com/gadgets/2025/04/google-used-ai-to-block-three-times-more-fraudulent-advertisers-in-2024/A New Zealand charity is using AI chatbots to help veterans: https://www.rnz.co.nz/news/national/558481/charity-launches-ai-chatbot-therapy-service-for-veteransOrganisations need to move strategically when rolling out AI, rather than rushing into it: https://www.informationweek.com/machine-learning-ai/the-ai-fomo-trap-build-guardrails-for-the-gold-rushAI hardware is using so much power that its driving advances in cooling technology for data centres: https://www.techrepublic.com/article/news-data-centers-power-cooling-ai/Google's AI is learning to speak dolphin: https://www.extremetech.com/science/new-google-llm-aims-to-translate-dolphin-languageMicrosoft's Copilot AI is getting quite annoying: https://www.theregister.com/2025/04/18/microsoft_copilot_not_wanted/A new approach to combat prompt injection attacks against generative AI: https://arstechnica.com/information-technology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/The key to employing AI effectively in marketing is to carefully train the model: https://dataconomy.com/2025/04/17/training-your-ai-not-just-your-team-a-marketers-guide-to-smarter-campaigns/The idea of 'adjusting' - read reducing - AI guardrails because a competitor has done so is rather frightening: https://techcrunch.com/2025/04/15/openai-says-it-may-adjust-its-safety-requirements-if-a-rival-lab-releases-high-risk-ai/Seeding AI with misinformation is being used more and more by bad actors: https://www.stuff.co.nz/world-news/360659194/russia-seeds-chatbots-lies-and-any-bad-actor-could-game-ai-same-wayWhile greater efficiencies in AI can allow more to be done for the same amount of energy, the demand is not necessarily there to do so: https://www.economist.com/finance-and-economics/2025/01/30/tech-tycoons-have-got-the-economics-of-ai-wrongI wouldn't describe it as 'thinking with pictures' as AI don't really 'think': https://www.computerworld.com/article/3964968/open-ais-new-models-can-think-with-pictures.htmlAI companies are fine with stealing other people's work, but hate it when someone does the same to them: https://www.theregister.com/2025/04/14/miyazaki_ai_and_intellectual_property/Could the use of screenshots help AI learn what is important to users? https://www.theverge.com/ai-artificial-intelligence/650809/screenshots-apps-ai-pixel-nothingA framework for constructing voice AI agents: https://www.kdnuggets.com/the-easiest-way-to-create-real-time-ai-voice-agents
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):
- How AI can help with taxes and financial planning: https://www.bigdatawire.com/2025/04/18/the-transformative-role-of-ai-in-financial-planning-and-tax-preparation/
- Actors who sold their likenesses for AI avatars are regretting doing so: https://arstechnica.com/ai/2025/04/regrets-actors-who-sold-ai-avatars-stuck-in-black-mirror-esque-dystopia/
- AI is widely used in the New Zealand public sector, mostly because it helps them with their work: https://www.rnz.co.nz/news/political/558081/ai-widely-used-in-public-sector-survey-finds
- One-bit neural network weights, to reduce the memory footprints of large AI: https://arstechnica.com/ai/2025/04/microsoft-researchers-create-super%e2%80%91efficient-ai-that-uses-up-to-96-less-energy/
- The role of Chief AI Officer is expected to become more prominent in coming years: https://www.informationweek.com/machine-learning-ai/how-will-the-role-of-chief-ai-officer-evolve-in-2025-
- How AI is affecting the creative professions: https://www.nzherald.co.nz/nz/how-ais-latest-advancements-could-reshape-creative-professions-the-front-page/NVIPGTLUR5GVFDBUN4FWJW23EE/
- AI agent hallucinates new policy, upsets users to the point of cancelling subscriptions: https://arstechnica.com/ai/2025/04/cursor-ai-support-bot-invents-fake-policy-and-triggers-user-uproar/
- AI models get more advanced, but they just hallucinate more: https://techcrunch.com/2025/04/18/openais-new-reasoning-ai-models-hallucinate-more/
- Wikipedia got so hammered by AI training bots scraping the site, they have released a data set optimised for training AI to Kaggle: https://www.theverge.com/news/650467/wikipedia-kaggle-partnership-ai-dataset-machine-learning
- Google is using AI to block scammers' ads: https://arstechnica.com/gadgets/2025/04/google-used-ai-to-block-three-times-more-fraudulent-advertisers-in-2024/
- A New Zealand charity is using AI chatbots to help veterans: https://www.rnz.co.nz/news/national/558481/charity-launches-ai-chatbot-therapy-service-for-veterans
- Organisations need to move strategically when rolling out AI, rather than rushing into it: https://www.informationweek.com/machine-learning-ai/the-ai-fomo-trap-build-guardrails-for-the-gold-rush
- AI hardware is using so much power that its driving advances in cooling technology for data centres: https://www.techrepublic.com/article/news-data-centers-power-cooling-ai/
- Google's AI is learning to speak dolphin: https://www.extremetech.com/science/new-google-llm-aims-to-translate-dolphin-language
- Microsoft's Copilot AI is getting quite annoying: https://www.theregister.com/2025/04/18/microsoft_copilot_not_wanted/
- A new approach to combat prompt injection attacks against generative AI: https://arstechnica.com/information-technology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/
- The key to employing AI effectively in marketing is to carefully train the model: https://dataconomy.com/2025/04/17/training-your-ai-not-just-your-team-a-marketers-guide-to-smarter-campaigns/
- The idea of 'adjusting' - read reducing - AI guardrails because a competitor has done so is rather frightening: https://techcrunch.com/2025/04/15/openai-says-it-may-adjust-its-safety-requirements-if-a-rival-lab-releases-high-risk-ai/
- Seeding AI with misinformation is being used more and more by bad actors: https://www.stuff.co.nz/world-news/360659194/russia-seeds-chatbots-lies-and-any-bad-actor-could-game-ai-same-way
- While greater efficiencies in AI can allow more to be done for the same amount of energy, the demand is not necessarily there to do so: https://www.economist.com/finance-and-economics/2025/01/30/tech-tycoons-have-got-the-economics-of-ai-wrong
- I wouldn't describe it as 'thinking with pictures' as AI don't really 'think': https://www.computerworld.com/article/3964968/open-ais-new-models-can-think-with-pictures.html
- AI companies are fine with stealing other people's work, but hate it when someone does the same to them: https://www.theregister.com/2025/04/14/miyazaki_ai_and_intellectual_property/
- Could the use of screenshots help AI learn what is important to users? https://www.theverge.com/ai-artificial-intelligence/650809/screenshots-apps-ai-pixel-nothing
- A framework for constructing voice AI agents: https://www.kdnuggets.com/the-easiest-way-to-create-real-time-ai-voice-agents
