AI Future — Frequently Asked Questions
Honest answers to the big questions about AI progress, timelines, risks, and what it means for ordinary people.
Frequently Asked Questions
How fast is AI actually improving?
Fast, but unevenly. Language and reasoning capabilities are improving roughly 2-3x per year on standard benchmarks. Image generation quality doubles roughly every 8 months. Robotics and physical-world AI are improving much more slowly — maybe 1.5x per year. The popular narrative of "exponential acceleration" is true for some capabilities and misleading for others.
When will we get AGI?
Depends entirely on your definition. If AGI means "an AI system that can perform any intellectual task a human can" — we're likely 5-10+ years away, and there are open theoretical questions about whether current approaches (transformers + scaling) get us there. If AGI means "AI that's very good at a wide range of tasks" — we're arguably already there.
The honest answer: nobody knows, and anyone who gives you a confident date is selling something.
Should I be worried about AI safety?
Yes, but about the right things:
| Real near-term risks | Overhyped risks |
|---|---|
| Deepfakes and misinformation | Killer robots |
| Job displacement without transition support | Sentient AI taking over |
| Privacy erosion through AI surveillance | AI "going rogue" |
| Concentration of AI power in few companies | Skynet scenarios |
| AI-assisted scams and social engineering | Existential risk (this decade) |
The risks that affect you in 2026 are mundane and real: getting scammed by an AI-generated voice clone, having your job restructured around AI tools you weren't trained on, or trusting AI output that's confidently wrong.
Will AI replace my job?
Probably not entirely. But it will almost certainly change what your job looks like. The pattern across history is clear: technology eliminates specific tasks faster than it eliminates entire jobs. ATMs didn't eliminate bank tellers — but tellers who adapted thrived while those who didn't struggled.
The jobs most affected first:
- Routine content creation (copywriting, basic design, data entry)
- Customer service (first-line support, FAQ handling)
- Administrative coordination (scheduling, email triage, reporting)
- Basic analysis (market research, competitive analysis, data summarisation)
How do I future-proof my career?
- Use AI tools in your current job. Become the person who shows your team how to use AI, not the person who resists it.
- Develop judgment skills. AI generates options; humans make decisions. Get better at evaluating, choosing, and taking responsibility.
- Build relationships. The more your work depends on trust, nuance, and human connection, the harder it is to automate.
- Learn to prompt well. Not as a career — as a meta-skill. Knowing how to get useful output from AI is the new literacy.
- Stay adaptable. The specific tools and platforms will change fast. The skill of learning new tools won't.
Is open-source AI catching up to commercial AI?
Yes, significantly. Llama 3 (Meta) and Mixtral (Mistral) are competitive with GPT-4 for many tasks. The gap between open-source and commercial frontier models has narrowed from ~18 months to ~6 months. For privacy-sensitive or cost-sensitive applications, open-source models running locally are increasingly the smart choice.
Last updated: March 2026