Maxim Levoshin

ru_RU Russian
ru_RU Russian

Category: AI

  • When ChatGPT booked me a haircut… in a bus

    When ChatGPT booked me a haircut… in a bus

    When ChatGPT booked me a haircut – and nailed it… sort of

    Today I decided to test ChatGPT’s new agent mode. I’d seen a post where it worked flawlessly: someone asked to book a haircut in Los Angeles, the AI opened a browser, found a salon, made the appointment – smooth as silk.

    Inspired, I tried: “Book me a haircut tomorrow in Brooklyn, São Paulo.”

    Spoiler: I did NOT expect what happened next.

    At first: flawless automation

    At first, it was perfect. The browser opened, and the agent narrated:
    “Comparing ratings,”
    “Scrolling the page,”
    “Clicking the booking button.”
    It found a barber, asked my preferred time, paused briefly for login, then finished the booking and sent me a neat summary: tomorrow at 1 PM, 25-minute haircut.

    I even got the email confirmation.

    It felt like the future — the kind where AI can do everything except find your missing socks.

    The twist: a barbershop in a bus

    But then… I checked the salon photos. And there it was: a barbershop. Inside. A regular. City. Bus.

    Not “retro-themed,” not “repurposed.” Just a literal bus parked on the roadside, with stools, dust, and a hint of existential despair.

    AI: mission accomplished.
    Me: wondering if my next agent prompt would be “Find where they sterilize their tools.”

    Lesson: trust but verify

    Always double-check — whether it’s a human agent or a digital one. Amen.

  • The Next 3 Years of AI: What to Expect and Why It Matters

    We’ve already passed the event horizon.
    The point of no return is behind us. While we haven’t yet built robots that handle every task or instant cures for disease, GPT-4 and other systems are already outperforming humans in many areas. And yes — they draw better cats. It’s only accelerating from here.

    Where AI is heading

    The most important shift isn’t just automation of routine work — it’s the acceleration of scientific progress and productivity. AI is already making researchers 2–3 times more productive.

    What’s coming (if current trends continue)

    2025: AI agents will perform complex cognitive tasks — such as writing code. Programming will be completely transformed.

    2026: We’ll see systems capable of generating new scientific ideas. Not just assisting — actually inventing.

    2027: Physical robots will begin entering the real world as useful agents.

    Intelligence will be cheaper than water

    By the 2030s, we’ll enter an era of abundance. The current barriers holding back progress will disappear. Intelligence will be as ubiquitous as electricity.

    Fun fact: a single ChatGPT query consumes just 0.34 Wh of energy.In the future, computing costs will be negligible.
    AI will help create new AI. Robots will build more robots. A cycle of self-improvement is underway.

    What this means for us

    Many jobs will disappear. New ones will emerge. Social systems will adapt — not instantly, but gradually. New social contracts and rules will be created.

    What seems like magic today — AI writing novels, developing new drugs, discovering new materials — will soon feel routine.
    Scientific breakthroughs in one year instead of decades.

    The biggest risks

    The single most critical challenge: solving the alignmentproblem — ensuring AI acts in humanity’s best interests.

    We already see how social media algorithms manipulate attention. If we don’t design future AI with care, the risks will be far greater than just the zombification of newsfeeds.

    This is a short summary of Sam Altman’s essay on the future of AI.

    My take?
    If you’re an entrepreneur or simply someone who wants to build new things — there has never been a better time to start.
    Well — except yesterday, of course.

    We’re standing at the beginning of a true world reset. And it’s only just getting started.