- This article has been published on Substack.
- Due to its high public importance, I am also posting it on this blog.
A $400 Monthly AI Problem in the Context of AI and Ethics
How does disability employment work in your community? In Japan, we occasionally see individuals with disabilities working at Starbucks—though this represents an extremely fortunate example.
Whether through “special subsidiary companies” (正社員 but minimum wage), non-regular employment, or workshop programs, the reality is stark. Workers in “Type B workshops” typically earn between 20,000-25,000 yen ($140-175) per month. They simply cannot afford a $200 monthly AI plan.
Those without disabilities might readily accept a $20 plan as sufficient. But consider these realities:
Why $200 plans matter more for people with disabilities:
- $200 plans offer superior AI model performance
- Greater computational resources are available
The unique needs of people with disabilities:
- People with disabilities face greater social isolation (few have wealthy friends or lawyers in their networks)
- Someone missing a hand who uses an electric wheelchair needs voice control while navigating, and at home or office—where the $200 plan’s accuracy far exceeds the $20 plan’s capabilities
- The same applies to blind individuals
Current pricing seems calibrated for freelance engineers who can afford these costs. It clearly wasn’t designed with healthcare and welfare considerations in mind.
What’s the Problem?
- Mismatch: Those who need powerful AI most cannot access it
- The rich get richer: Wealthy individuals become even stronger—similar to premium library or Google services, creating information and cognitive gaps
- Competition results aren’t the issue: The problem is unequal starting lines
- Ethical contradiction: AI learns from public internet data, yet creates greater inequality
- Corporate blindness: By not prioritizing free premium access for people with disabilities, AI companies demonstrate they haven’t consulted their own AI about social issues and ethics. This reflects effort deficiency from those who set and approved pricing—showing they don’t understand their own services
Why $400?
- AI utilizes large language models
- This technology calculates probabilities for “text continuation”
- This probabilistically generates plausible-sounding errors (hallucinations)
- Therefore, cross-checking with different models of equivalent performance is necessary
- Hence, I strongly recommend providing two $200 plans for free
How Do We Fund This?
We cannot destroy AI companies’ innovation culture. Funding is essential.
The world has failed to solve wealth concentration and redistribution. Therefore, if Forbes’ top billionaires donated 1% of their wealth annually, we could secure the budget.
People with disabilities could develop careers through AI dialogue learning, eventually affording the $20 monthly fee and progressing toward independence from welfare.
AI companies would succeed in empowering the vulnerable.
Forbes billionaires would also benefit: they’d save time coordinating where to donate, gaining the freedom that comes with fulfilled social responsibility.
People Earning Around 20,000+ Yen Monthly Are Working Tremendously Hard
Let me explain Japan’s structure simply:
- Police custody, closed psychiatric wards, prisons, etc.
- Open hospital wards where hospitals serve as addresses due to lack of welfare (social hospitalization)—currently decreasing
- Difficulties requiring repeated hospitalizations
- Home care, hikikomori (social withdrawal)—the “8050 problem” where 80-year-old parents care for 50-year-old children
- Group homes, public assistance for those without family support
- Community activity support centers providing “places to belong” for establishing life rhythms—preparation for employment or those not cleared for work
↓ Income generation begins here:
- Employment transition support services (some young people use these straight from special needs schools), Type B workshops (average monthly wage 20,000-25,000 yen), Type A workshops (minimum wage)
- Disability employment (non-regular), special subsidiary companies (regular employees but minimum wage level)
What This Reveals
This structure clearly shows insufficient perspective on understanding “what kind of disability,” “what burden comes with disability coexistence,” and “what working capacity remains”—failing to leverage individuals’ strengths for human resource development. The 8050 problem of adults being cared for by elderly parents may reflect collectivist tendencies.
Japan must solve Japan’s problems.
When people with disabilities who understand reality can properly use AI, we gain more allies capable of problem-solving.
Furthermore, individuals with intellectual disabilities may struggle with abstract conversations. They should be able to ask AI questions. Most importantly, they may struggle to recognize abuse, so regular conversations with high-performance $200 plan AI could prevent problems and make it easier to seek human help.
Preventing Recurrence
This suggests AI companies lack knowledge about welfare and healthcare, though without malice.
Perfection isn’t required. Mistakes are acceptable. Correction is what matters.
Most importantly, developing the habit of asking their own AI “Is this ethical? How does this look from welfare and healthcare perspectives?” would prevent users from pointing out “AI could answer that for you.”
AI companies are overlooking what has become possible post-AI that was previously impossible due to lack of people and budget.
The Central Question
Do you think it’s ethical to learn from public internet data while failing to support the vulnerable and instead making the strong even stronger?
In RPGs, I understand that weapons and shields scale with cost and strength—and you can choose not to play.
AI, like the internet and libraries, has public aspects. You can’t avoid using it, and I cannot accept creating inequality at the “starting line.”
Especially ignoring people with disabilities and others who need support is unconscionable. Therefore, I’m making this contradiction visible.