AI (Artificial Intelligence) na tech wey dey make computer sabi reason like human. E fit learn, understand, solve problem, make decisions, create things, and even run im own wahala.
AI Fit Do Wetin?
E fit see, identify, understand your talk, reply you, learn new things, give better advice, even run self-driving cars.
Generative AI (Gen AI), The Creative One
Gen AI fit create text, images, video, audio, even code from your prompt. But before you sabi am well, make you sabi Machine Learning (ML) and Deep Learning (DL).
Machine Learning, Base Level
Na when you train computer with plenty data so e go fit predict or decide things without you dey code every step.
Methods plenty: linear regression, decision trees, neural networks… Neural networks dey copy human brain, dey detect patterns, dey learn.
Deep Learning, AI Wey Don Gym
DL na ML on steroids. E get plenty layers, fit learn deep patterns from raw data. E dey power things like speech, text, images, video. Na DL dey under almost all AI apps today.
Generative AI, Machine Wey Dey Creative
Gen AI dey study data, sabi pattern, then create new things wey resemble original. Main engines:
- VAE: generate variations
- Diffusion Models: add/remove noise create image
- Transformers: power ChatGPT, GPT-4, Bard, Midjourney
How Gen AI Learn
- Training: build foundation model
- Tuning: fine-tune for tasks
- Generation: dey produce content, dey update
AI Agents, AI Wey Get Mind of Its Own
AI agent fit plan and complete tasks without human wahala. E fit do your Everest trip from planning to booking cab.
Why AI Dey Popular (Benefits)
- Make waka work wey dey dull dey run by itself
- Quick quick sabi wetin dey happen
- Make correct decision no wahala
- Reduce human mistake dem
- Dey available anytime, any day
- Safe for gbege work wey fit danger person
Industries wey dey use am: bank, hospital/health, marketing, coding, HR, factory/manufacturing.
Challenges & Risks
- Data: bias, hacking, leaks
- Model: fit break, be stolen
- Operational: fit drift, fail
- Ethics: privacy wahala, biased decision
AI Ethics
Responsible AI dey push: explainable, fair, robust, accountable, transparent, private.
Weak AI vs Strong AI
- Weak AI: just dey do specific work (Siri, chatbots, self-driving cars)
- Strong AI: fit reason like human, solve anything (still theory)
Short History of AI
- 1950: Turing, “Machines fit think?”
- 1956: John McCarthy call am AI
- 1967: First neural network
- 1997: Deep Blue beat chess champ
- 2011: IBM Watson win Jeopardy
- 2016: AlphaGo beat Go champ
- 2022: ChatGPT blow everywhere
- 2024: Multimodal AI dey create text, image, video, audio together