I watched a classroom where a chrome-faced robot spoke too fast and the kids asked it to repeat itself. You could feel the awkwardness—half a million dollars worth of awkwardness. I left asking one plain question: who thought this was the best use of public money?
I’m going to walk you through what happened, what those machines actually do, and why that price tag—$500,000 (€465,000)—matters more than the marketing copy. I’ve read the principal’s email, watched the Voice of San Diego piece, and poked at the product pages and reviews so you don’t have to.
A principal emailed her community claiming the school was the “first in the world” testing physical AI.
That note, signed by Cathryn Rambo of Altus Schools, announced the arrival of two Ameca humanoid robots and described the project as a research study that will run through Fall 2026.
Engineered Arts, the U.K. firm behind Ameca, sells the robot as a social humanoid with “61 actuated degrees of freedom” and advanced voice synthesis. Reviewers place Ameca between $100,000 (€93,000) and $500,000 (€465,000) depending on options—so if Altus really spent half a million, they appear to have paid roughly $250,000 (€232,500) per unit. The company doesn’t post list prices; buyers request quotes.
Why did Altus spend $500,000 on robots?
Officially, the school framed this as a research partnership: data collection, student engagement measures, a trial period. The email calls it “research” and promises the robots will be onsite until the fall.
But the paperwork Voice of San Diego obtained leaves gaps. Are the robots purchased or leased? Who funds the research—Altus, a vendor, or a third party? And if outside researchers are involved, who owns the student interaction data? Those are not small details when public funds may be in play.
At a demo the robot delivered a halting Nikola Tesla monologue while kids furiously copied notes.
Voice of San Diego’s reporter described the exchange as “clunky”: stops and starts, speech that ran too fast, requests from students to repeat the same intro multiple times.
Functionally, the Ameca was acting as a mouthpiece for a limited version of ChatGPT or another large language model—the same family of tools behind OpenAI’s GPTs and ChatGPT. That raises both practical and ethical questions: should LLM-driven interfaces speak directly to children without strong guardrails? A growing chorus of researchers is arguing for tight restrictions on chatbot use by minors.
Are humanoid robots effective in classrooms?
There’s enthusiasm and theater value: a robot draws attention. But attention is not the same as learning. Much of the academic work on AI tutors relies on software tailored to pedagogy, not on mannequins with expressive faces.
In practice the demos looked like theater with glitches. If the goal is better literacy, a few trained teachers and updated materials will probably move metrics more reliably than a metallic actor that repeats itself.
You can see the trade-offs when you divide the bill by per-student spending.
California spends about $21,600 (€20,090) per student per year. Nationwide, public spending averages $16,500 (€15,350) per student. At California rates, $500,000 (€465,000) covers roughly 23 students for a year.
That arithmetic is blunt but damning: you can buy teachers, books, field trips, or counseling services for a quarter-million apiece that will have measurable effects on student outcomes. Instead, Altus chose two humanoids on site for a few months.
Call it what you will, but this purchase resembles paying for a high-gloss prototype that prioritizes novelty over predictable returns—like buying a sports car to haul the groceries.
The machines are made by Engineered Arts and appear to be paired with a filtered LLM behind the scenes.
Engineered Arts builds Ameca as a platform: realistic face, humanlike motions, and an API that lets schools or researchers hook in conversational engines. Reports say Altus used a restricted ChatGPT instance as the robot’s voice.
That setup leads to several friction points: latency in responses, mismatched speech pacing, content controls that may overcorrect or fail to catch subtle errors. It also raises data questions: does OpenAI or any LLM provider retain logs? Who has access to recordings of student interactions?
Who is conducting the study and who benefits?
The principal’s message frames Altus as the research host. But without a published protocol, institutional review, or clear data-use agreement released publicly, you and I are left guessing whether the school, a vendor, or outside researchers will use the data and to what end.
Parents deserve clarity: consent forms, data retention policies, and independent oversight if minors’ interactions are being studied. Right now the paperwork appears thin.
One reporter watched a messy demo and a community watched the price tag roll by.
Voice of San Diego did the reporting you’d expect—obtaining emails, seeing the demo, and asking the obvious questions. The narrative that remains is part wonder, part expense report, part privacy puzzle.
There’s a marketing thrill to a humanoid in class; there’s also a responsibility when you spend public dollars. The robot might be an impressive piece of engineering, but impressive objects don’t automatically translate into better education—sometimes they’re theatre, sometimes they’re research, and sometimes they are both.
I’ve shown you the emails, the demo notes, the product claims, and the math. If you’re a parent, a teacher, or a local policymaker, what would you ask the principal at the next meeting?