Oslo acquires ChatGPT for 110k students and teachers
Oslo’s decision to buy GPT‑3.5 access for 110,000 students and teachers is held up as a pragmatic move that contrasts sharply with schools elsewhere that try to ban generative AI. Commenters debate whether giving children institutional access to ChatGPT will narrow or widen inequalities (e.g., versus students who can pay for GPT‑4), how to handle privacy and data protection, and whether open‑source models should have been chosen instead. A recurring theme is whether AI in classrooms will erode fundamental skills and critical thinking or, if used carefully, become as essential and beneficial as calculators and search engines once were.
Overall Reaction & Context
- Many contrast Oslo’s move with US school bans and Italy’s temporary ban, seeing it as a pragmatic acceptance that LLMs are here to stay.
- Some view it as potentially “genius” early adoption that may better prepare students for a future with ubiquitous AI assistants.
- Others see it as a performative, trend-chasing decision by officials, similar to underplanned rollouts of iPads or school laptops.
Model Choice, Licensing, and Privacy
- Deal is reportedly GPT‑3.5 via Azure / enterprise setup, with terms that student data won’t be used for training.
- Several posters note this likely addresses EU child‑data and age‑limit issues and justifies paying instead of using free public ChatGPT or Microsoft Copilot.
- Some confusion over whether GPT‑4 is included for some institutions; overall, GPT‑4 access is described as more expensive.
Educational Goals and Teacher Roles
- Supporters argue schools should teach AI literacy, including prompt writing, limitations, and ethical use, rather than pretend students aren’t using it.
- Debate over teacher roles: some predict many teachers could be replaced by AI plus a few supervisors; others strongly reject this, stressing teachers’ roles in mentoring, social development, and emotional support.
Quality, Hallucinations, and Trust
- Strong thread on hallucinations: many distrust ChatGPT as a “lie box” that confidently presents fiction as fact.
- Others argue it’s valuable if treated as a fallible assistant, especially where answers are easy to verify (coding, math steps, scripts) or where correctness is less critical (summarization, creative writing).
- Some hope formal use in schools could explicitly teach skepticism and fact‑checking; others think that giving kids an authoritative but unreliable tool is irresponsible.
Equity, Access, and Inequality
- Concern that students with private GPT‑4 subscriptions gain an advantage over those limited to GPT‑3.5.
- Counterpoint: 3.5 is “good enough” for learning basic use; the bigger divide is between using AI effectively vs. not at all.
Open Source vs. Proprietary Options
- Several suggest Mixtral/Mistral or LLaMA‑based systems as better or more transparent than GPT‑3.5.
- Pushback notes: running large open models at Oslo scale needs GPUs and significant infrastructure, plus custom UI and account management; paying OpenAI/Microsoft is simpler and aligns with current “industry standard” tools.
Impact on Students’ Skills and Development
- Worries that heavy AI use will weaken students’ ability to research, memorize, and critically analyze information, making them more manipulable.
- Others compare this to past fears about calculators, GPS, writing, radio, and TV; they expect a shift in which skills matter (e.g., synthesis and cross‑disciplinary judgment) rather than a blanket decline in ability.