# Let’s be honest, Generative AI isn’t going all that well **By:** Gary Marcus **Published:** 2026-01-12T00:00:00+00:00 **Source:** [Substack](https://garymarcus.substack.com/p/lets-be-honest-generative-ai-isnt?r=8tdk6&triedRedirect=true) --- ### A sampling of recent news Some recent news, all long anticipated by this newsletter: 1. LLMs can still cannot be trusted: [ ![](https://substackcdn.com/image/fetch/$s_!J1d8!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F36114186-2f6a-4209-b8f7-96b2cc304900_730x1032.png) ](https://substackcdn.com/image/fetch/$s_!J1d8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F36114186-2f6a-4209-b8f7-96b2cc304900_730x1032.png) [Full article](https://www.theguardian.com/technology/2026/jan/11/google-ai-overviews-health-guardian-investigation?CMP=Share_iOSApp_Other) 1. A large fraction of what LLMs do is mostly just memorization (and Hinton was on the wrong side of [this argument](https://open.substack.com/pub/garymarcus/p/deconstructing-geoffrey-hintons-weakest?r=8tdk6&utm_medium=ios&shareImageVariant=overlay))_:_ [ ![](https://substackcdn.com/image/fetch/$s_!kt8t!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe4e5621d-c50c-482c-9d4a-ab6dff8f8a24_2401x1125.png) ](https://substackcdn.com/image/fetch/$s_!kt8t!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe4e5621d-c50c-482c-9d4a-ab6dff8f8a24_2401x1125.png) [Full article here.](https://www.theatlantic.com/technology/2026/01/ai-memorization-research/685552/?gift=NJpHw9fdcmQdiPa0tCBLEkPpq6_lwPrX-9Ktk1-uZPY&utm_source=copy-link&utm_medium=social&utm_campaign=share) 1. They still aren’t adding a lot of quantifiable value to the world: [ ![](https://substackcdn.com/image/fetch/$s_!Vc9J!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8c6f67d7-fc7d-4bbd-a881-42aa4f5d35df_900x1314.png) ](https://substackcdn.com/image/fetch/$s_!Vc9J!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8c6f67d7-fc7d-4bbd-a881-42aa4f5d35df_900x1314.png) via newsletter _Update: This is consistent with the finding of the [Remote Labor Index](https://arxiv.org/abs/2510.26787) that AI could only do about 2.5% of jobs, [reported recently by the Washington Post](https://www.washingtonpost.com/technology/interactive/2026/ai-jobs-automation/)._ 1. Scaling isn’t going all that well, anymore, and probably won’t cure these problems. [ ![](https://substackcdn.com/image/fetch/$s_!zrPY!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3f678a8-7b07-464f-8a34-de7edc9b9b1c_1547x1271.png) ](https://substackcdn.com/image/fetch/$s_!zrPY!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3f678a8-7b07-464f-8a34-de7edc9b9b1c_1547x1271.png) [Full article](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5877662) Trying to orient our economy and geopolitical policy around such shoddy technology — particularly on the unproven hopes that it will dramatically improve– is a mistake. The fundamental problem with generative AI is that the creators overpromised. The underlying technology itself is fine. There is a real business here. It is just not what the creators are saying it is. It is substantially smaller and less significant. dang I knew i had seen something else recently that i meant to include. now added as an update on #3.