Weekly Review 31 January 2025
Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):How to deploy a machine learning model to production: https://www.kdnuggets.com/guide-deploying-machine-learning-models-productionAI companies walk a fine line when dealing with the military: https://techcrunch.com/2025/01/19/the-pentagon-says-ai-is-speeding-up-its-kill-chain/AI is one technology that will be used by UK retailers to reduce staff costs: https://www.theguardian.com/business/2025/jan/21/robot-packers-and-ai-cameras-uk-retail-embraces-automation-to-cut-staff-costsYann LeCun predicts that the current large language model of AI will be superseded in five years: https://techcrunch.com/2025/01/23/metas-yann-lecun-predicts-a-new-ai-architectures-paradigm-within-5-years-and-decade-of-robotics/Tech companies are shovelling AI into everything. Shouldn't consumers have a choice? https://www.theregister.com/2025/01/23/why_is_ai_optout/Using AI to detect potholes. Now use AI to fix them! https://www.theregister.com/2025/01/21/ai_pothole_patrol/A lot of businesses that invested in AI will struggle to get a return on value: https://www.theregister.com/2025/01/22/business_value_genai_elusive/Companies are still struggling to find value in rolling out AI: https://www.informationweek.com/machine-learning-ai/why-enterprises-struggle-to-drive-value-with-aiDriverless buses is the best use of autonomous AI in transportation: https://spectrum.ieee.org/driverless-busThe first AI software engineer is really bad at its job: https://www.theregister.com/2025/01/23/ai_developer_devin_poor_reviews/You know things are getting bad when even the Pope is criticising AI: https://www.theguardian.com/technology/2025/jan/23/pope-warns-davos-summit-that-ai-could-worsen-crisis-of-truthThe real issue with sentient AI is how we would treat it: https://spectrum.ieee.org/sentient-aiGame companies are pushing their employees to use AI, despite employees' misgivings: https://www.extremetech.com/gaming/game-developers-are-worried-about-ai-but-studios-are-using-it-anywayIf the company board is reluctant to pursue AI, how can they be persuaded? https://www.informationweek.com/it-leadership/how-to-persuade-an-ai-reluctant-board-to-embrace-critical-changeLinkedIn has been allowing access to its members private messages for training AI: https://www.theregister.com/2025/01/22/linkedin_sued_for_allegedly_training/How good is AI at selecting someone's outfit? https://www.nzherald.co.nz/lifestyle/what-shall-i-wear-i-trusted-ai-to-dress-me-it-was-really-ageist/TCYC5MVYGFGRBGZ4U7F63J6Y64/If there's too many workers with degrees, it's because successive governments have messed up the funding of universities: https://www.thepost.co.nz/nz-news/360553238/killing-productivity-degreesAI can be useful in finding mathematical proofs, if they are used carefully: https://www.datasciencecentral.com/using-ai-to-solve-the-deepest-math-conjectures/Some of the data biases that affect AI and how to avoid them: https://www.datasciencecentral.com/combating-bias-in-ai-systems-a-guide-for-data-scientists/AI suck at history. Not surprising, they have no grasp of reality: https://www.computerworld.com/article/3806840/todays-ai-models-have-a-poor-grasp-of-world-history.htmlA cavalcade of AI-bungled news summaries shows that AI still has a long way to go before it can be relied upon: https://www.computerworld.com/article/3806774/apple-is-the-latest-company-to-get-pwned-by-ai.htmlAs we become more dependent on AI, we will be more impacted by outages: https://www.theregister.com/2025/01/23/chatgpt_has_a_thursday_lie/Great, another way to use AI to discriminate against people: https://www.computerworld.com/article/3804964/ai-can-predict-career-success-from-a-facial-image-study-finds.htmlAn AI agent that can do things for you is probably more useful, although I'm not keen on it having too much control of my computer: https://arstechnica.com/ai/2025/01/openai-launches-operator-an-ai-agent-that-can-operate-your-computer/
- How to deploy a machine learning model to production: https://www.kdnuggets.com/guide-deploying-machine-learning-models-production
- AI companies walk a fine line when dealing with the military: https://techcrunch.com/2025/01/19/the-pentagon-says-ai-is-speeding-up-its-kill-chain/
- AI is one technology that will be used by UK retailers to reduce staff costs: https://www.theguardian.com/business/2025/jan/21/robot-packers-and-ai-cameras-uk-retail-embraces-automation-to-cut-staff-costs
- Yann LeCun predicts that the current large language model of AI will be superseded in five years: https://techcrunch.com/2025/01/23/metas-yann-lecun-predicts-a-new-ai-architectures-paradigm-within-5-years-and-decade-of-robotics/
- Tech companies are shovelling AI into everything. Shouldn't consumers have a choice? https://www.theregister.com/2025/01/23/why_is_ai_optout/
- Using AI to detect potholes. Now use AI to fix them! https://www.theregister.com/2025/01/21/ai_pothole_patrol/
- A lot of businesses that invested in AI will struggle to get a return on value: https://www.theregister.com/2025/01/22/business_value_genai_elusive/
- Companies are still struggling to find value in rolling out AI: https://www.informationweek.com/machine-learning-ai/why-enterprises-struggle-to-drive-value-with-ai
- Driverless buses is the best use of autonomous AI in transportation: https://spectrum.ieee.org/driverless-bus
- The first AI software engineer is really bad at its job: https://www.theregister.com/2025/01/23/ai_developer_devin_poor_reviews/
- You know things are getting bad when even the Pope is criticising AI: https://www.theguardian.com/technology/2025/jan/23/pope-warns-davos-summit-that-ai-could-worsen-crisis-of-truth
- The real issue with sentient AI is how we would treat it: https://spectrum.ieee.org/sentient-ai
- Game companies are pushing their employees to use AI, despite employees' misgivings: https://www.extremetech.com/gaming/game-developers-are-worried-about-ai-but-studios-are-using-it-anyway
- If the company board is reluctant to pursue AI, how can they be persuaded? https://www.informationweek.com/it-leadership/how-to-persuade-an-ai-reluctant-board-to-embrace-critical-change
- LinkedIn has been allowing access to its members private messages for training AI: https://www.theregister.com/2025/01/22/linkedin_sued_for_allegedly_training/
- How good is AI at selecting someone's outfit? https://www.nzherald.co.nz/lifestyle/what-shall-i-wear-i-trusted-ai-to-dress-me-it-was-really-ageist/TCYC5MVYGFGRBGZ4U7F63J6Y64/
- If there's too many workers with degrees, it's because successive governments have messed up the funding of universities: https://www.thepost.co.nz/nz-news/360553238/killing-productivity-degrees
- AI can be useful in finding mathematical proofs, if they are used carefully: https://www.datasciencecentral.com/using-ai-to-solve-the-deepest-math-conjectures/
- Some of the data biases that affect AI and how to avoid them: https://www.datasciencecentral.com/combating-bias-in-ai-systems-a-guide-for-data-scientists/
- AI suck at history. Not surprising, they have no grasp of reality: https://www.computerworld.com/article/3806840/todays-ai-models-have-a-poor-grasp-of-world-history.html
- A cavalcade of AI-bungled news summaries shows that AI still has a long way to go before it can be relied upon: https://www.computerworld.com/article/3806774/apple-is-the-latest-company-to-get-pwned-by-ai.html
- As we become more dependent on AI, we will be more impacted by outages: https://www.theregister.com/2025/01/23/chatgpt_has_a_thursday_lie/
- Great, another way to use AI to discriminate against people: https://www.computerworld.com/article/3804964/ai-can-predict-career-success-from-a-facial-image-study-finds.html
- An AI agent that can do things for you is probably more useful, although I'm not keen on it having too much control of my computer: https://arstechnica.com/ai/2025/01/openai-launches-operator-an-ai-agent-that-can-operate-your-computer/