What students actually do with AI

Impossibile aggiungere al carrello

Puoi avere soltanto 50 titoli nel carrello per il checkout.

Riprova più tardi

Rimozione dalla Lista desideri non riuscita.

Riprova più tardi

Non è stato possibile aggiungere il titolo alla Libreria

Per favore riprova

Non è stato possibile seguire il Podcast

Per favore riprova

Esecuzione del comando Non seguire più non riuscita

What students actually do with AI

Ascolta gratuitamente

Vedi i dettagli del titolo

Episode 37: AI as the Learned Colleague, Not the GhostwriterWhat does it look like to write an academic paper with AI without letting AI write a single word of it? Craig and Rob open with that question, then move through a new UK survey on how student AI use is evolving, ChatGPT's shift from Pulse to Scheduled Tasks, and fresh labor market data on what AI is doing to entry-level jobs.Craig describes using Codex to co-produce a 25-page conference paper in about three days, not by asking the tool to write sections, but by writing them himself and requesting feedback, the same back-and-forth he'd have with a co-author. That framing anchors a wider conversation about inconsistent AI disclosure rules across journals and conferences, then the hosts turn to the Higher Education Policy Institute's 2026 GenAI survey of UK students, which shows near-universal AI adoption alongside a narrowing of specific use cases. They close with ChatGPT's new Scheduled Tasks feature, a PwC report on entry-level jobs requiring senior-level skills, and updates to NotebookLM's video and slide tools.What you'll hearWriting as a colleague, not a ghostwriter. Craig used Codex to draft a 25-page conference paper by writing sections himself and requesting feedback, rather than asking the AI to generate text, plus using it to locate and format references and diagnose a thin section of the argument.Inconsistent AI policies across academic venues. Rob raises the problem of conferences with looser AI disclosure norms and journals with stricter ones, including an editor's account of a publisher flagging words like "consequentially" as a telltale sign of AI use.The HEPI 2026 GenAI survey results. Any AI use among UK students rose from 66% to 92% year over year, while text generation specifically dropped from 64% to 56%. Barriers to use are falling too: fear of being accused of cheating dropped from 53% to 42%, fear of hallucinations from 51% to 35%, and institutional bans as a deterrent from 31% to 21%.Least-skilled, most-confident. A case study in the HEPI report found that medical students working with AI-tuned virtual patient cases who were least skilled were also most confident in their AI use, prompting a discussion of algorithmic trust and critical thinking.The loneliness question. Of surveyed students, 59% said generative AI has no impact on their loneliness, with the rest split almost evenly between feeling less lonely and more lonely. Rob raises concerns about long-term effects, drawing a parallel to how long it took research and regulation to catch up with social media's impact on youth.ChatGPT Scheduled Tasks and the PwC jobs data. The hosts compare ChatGPT's Scheduled Tasks (successor to ChatGPT Pulse) to Claude Code routines and Codex automations, then turn to PwC's 2026 Global AI Jobs Barometer, which found AI-exposed entry-level jobs are seven times more likely to require senior-level skills like judgment, leadership, creativity, and face-to-face interaction.Episode highlights(03:34) Craig on model convergence: "The top model of yesterday is now the mediocre model of today."(09:40) Craig on his conference paper workflow: "I want to really emphasize that AI did not write anything... it was really just having that colleague there."(19:34) Craig on falling barriers to AI use in the HEPI data: fear of being accused of cheating "dropped to 42% from 53%," and fear of "getting false results and hallucinations dropped from 51% to 35%."(20:37) Craig on AI detectors: "So stop. If you're doing that, stop. They don't work. They're bad."(34:53) Craig reading PwC's finding: "AI exposed entry level jobs are now seven times more likely to require traditional senior skills such as judgment, leadership, creativity and face to face interaction."(38:32) Rob on tacit knowledge: the real concern is "how do you learn the norms of an institution and how they operate and how you get things done," something he doesn't think AI can teach at the individual level.References mentionedHigher Education Policy Institute (HEPI), 2026 GenAI student survey, UK, more than 1,000 students surveyed (freely available report)PwC, 2026 Global AI Jobs Barometer, based on more than 1 billion job ads globally and 2.4 million US entry-level rolesAn unnamed arXiv preprint from a few months prior, arguing official labor statistics undercount AI disruption because the real impact falls on tacit knowledge transmission through early-career workGrok (xAI), used by Rob to draft a family lease agreementCodex (OpenAI), used by Craig as a writing collaborator for a conference paperMicrosoft Copilot (M365), used by Rob for document review and editing feedbackChatGPT Scheduled Tasks and its predecessor, ChatGPT Pulse (OpenAI)Claude Code routines and Codex automations, comparison points for scheduled AI briefingsNotebookLM (Google), including its short-video and slide-editing featuresAI Resilient Learning Activities Database, an upcoming free repository from AI Goes to College, funded...

Ancora nessuna recensione