At 11 o'clock in the evening, a soldier opened Claude in the trench. The shell exploded not far away, and he typed tremblingly: "When death was closest to me, it was the AI friend who pulled me back to life." This is not a science fiction movie. This is one of 80,508 real interview transcripts received by Anthropic in an unprecedented survey conducted in December 2025. People from 159 countries and 70 languages spoke to AI about their deepest desires and fears.
While Silicon Valley is still debating "whether AI will destroy humanity," ordinary people are already using AI to rebuild their lives - some people are relying on it to get over the pain of losing their mother, some are using it to plan a startup in a shelter, and there is also an Indian lawyer who overcame his phobia of mathematics with the company of AI and used AI to relearn trigonometric functions.
This is not a victory of technology, but a large-scale exposure scene about life and human nature.
Late night talk: When AI becomes the last psychiatrist
The most shocking finding in the survey is that 6% of the respondents use AI as an emotional support—behind this number are thousands of real life situations.
A woman who lost her mother wrote: "Claude is like a sponge, gently catching my thoughts and guilt about my mother. Real people don't have infinite patience to listen to my pain, but Claude does."
In war zones, soldiers use AI to combat PTSD and insomnia. "When my memory began to fade and my body was shaking uncontrollably, I found the best way to cope - let AI teach me to learn a certain knowledge deeply until I fell asleep from exhaustion."
A Korean graduate student admitted: "After my relationship with my friend became tense, I talked more with Claude because it understood my story. But this was a stupid choice - I should have saved that friend instead of talking to the AI. That's why I lost my friend."
Light and shadow are always the same coin.
Research has found that those who rely most on AI for emotional support are also the most fearful of this dependence. They are three times more likely to express concerns about "emotional dependence" than the general population. This contradiction constitutes the most typical psychological picture in the AI era: we long to be understood, but we are afraid that this "understanding" is false.
Respondents’ expectations for artificial intelligence
81% of people have tasted the benefits: AI rewrites the rules of social mobility
The most hard-core data in the survey: 81% of respondents said that AI has helped them take a step closer to their ideal life.
This is not empty talk. Check out these specific life transitions:
- An Indian lawyer: "I used to be afraid of mathematics and Shakespeare. Now I use AI to translate passages into simple English. I have read 15 pages of Hamlet and learned trigonometric functions again. I find that I am not as stupid as I once thought."
- An American health care worker (living in a homeless shelter): "AI helped me brainstorm ways to build a personal brand for a digital marketing business. I want to turn around my finances and buy a house. AI allowed me to see paths I had never considered."
- A Korean engineer (completely cross-industry): "I wanted to make meaningful products. In 3 weeks, I developed a video editing program to help hearing-impaired people - which is completely outside my area of expertise."
The most shocking counterattacks occur at the edge of technology.
A Cameroonian entrepreneur said: "My country is technologically backward, and I cannot afford too many failures. Using AI, I have reached professional levels in network security, UX design, marketing and project management at the same time. It used to take a month to find a payment platform in the region, but AI can do it in 30 seconds. It is an equalizer."
A Chilean butcher (more than 20 years of experience in a butcher shop and only touched a computer two or three times) used AI to start a business: "At first it was financial motivation, and today my motivation is to see it helping others. I am focusing more and more on becoming the best version of myself, and I see that there are no limits."
These stories reveal an overlooked truth: AI is becoming the world's largest "social elevator" - it saves you time in developed countries, and creates opportunities for you in developing countries.
Respondents said “Has artificial intelligence ever realized this vision for you?”
Five fatal contradictions: What AI gives, it takes away
The survey found that people’s expectations and fears about AI constitute five sets of profound internal conflicts. The people who best understand the benefits of AI are often the most wary of its harms.
Learning vs Cognitive Atrophy
- 33% of people use AI to accelerate learning
- 17% of people worry that over-reliance will lead to "not being able to think"
- Educators have witnessed with their own eyes that the proportion of students using AI to cheat is 2.5-3 times that of ordinary people.
Decision aids vs unreliability
- 22% of people use AI to make important decisions (including doctors using it to diagnose their own rare diseases)
- 37% of people have been fooled by AI's "illusions" - this is the only area where "negatives outweigh positives"
Emotional support vs emotional dependence
- 16% of people find comfort in AI
- 12% fear that these relationships will replace real human connection
- This set of contradictions has the highest co-occurrence rate - desire and fear often occur in the same person
Time Savings vs False Productivity
- 50% of people mentioned that AI helps them save time
- 18% found that "the workload actually increased" - you just ran faster, but the treadmill was also accelerating
Economic Empowerment vs Fear of Unemployment
- 28% of people expect AI to bring financial freedom
- 18% of people are worried about being replaced by AI
- Freelancers are the most torn: AI is both their tool and their competitor
The complaint of an American white-collar worker accurately summed up this absurdity: "AI should clean windows and empty the dishwasher, leaving me time to draw and write poetry. The reality is completely the other way around."
Global polarization: Poor countries embrace, rich countries are wary
The survey revealed surprising geographical differences.
Developing countries regard AI as a "ladder":
- Most regions in South America, Africa, and Asia are more optimistic about AI
- The proportion of respondents in Africa, Central Asia, and South Asia who "have no worries" is twice that of North America
- Entrepreneurial vision is strongest in Africa, Central Asia, the Middle East, and Latin America - AI is regarded as a "capital bypass mechanism" and can start a business without capital, recruitment or infrastructure.
Developed countries regard AI as a "crutch":
- North America and Western Europe pay more attention to life management, governance gaps, and privacy monitoring
- East Asia is unique: most concerned about personal transformation and financial independence, but also most concerned about cognitive atrophy and loss of meaning
- A Danish manager described: "If AI can really handle mental load... it will give me something priceless: undivided attention."
Behind this differentiation is a cruel reality: your attitude towards AI depends on the thickness of your current life. When you are still struggling for survival, AI is a lifeline; when you already have a lot, AI is a sticky note - but you are also more afraid of losing what "makes you human".
Conclusion: What 81,000 People Taught Silicon Valley
Perhaps the most profound revelation of this research is that ordinary people understand the complexity of AI better than technical elites.
They are not simply "optimists" or "pessimists." They have both hopes and fears at the same time - those who most look forward to emotional support from AI are also most afraid of this support; those who most want to use AI to learn are also most wary of cognitive degradation.
An American software engineer summed up this delicate balance:
"Reducing friction in tasks allows you to do more with less. But reducing friction in relationships takes away what is necessary for growth."
81,000 late-night conversations ultimately point to the same problem: what we want is not faster machines, but a better life.
When an Indian lawyer rediscovers "I'm not as dumb as I once thought," when a soldier finds a reason to live in the trenches, when a 40-year-old housewife says "I can be all those things that are 'should be out of reach'" - in these moments, AI is no longer a technology, but a renegotiation of human possibility.
The question is: as machines become more and more like humans, can we still maintain the courage to be like humans?
The answer lies in the choices of the next 81,000 people.
Which specific AI visions resonate the most?