How does AI actually work, and is it really intelligent?

In episode 3 of No Fixed Lane, Rashad and Mammad break down what happens under the hood of today's AI: tokenization, embeddings, latent space, and next-token prediction. They explain why current models are better understood as powerful computation plus smart algorithms than as human-like intelligence.

The conversation moves into real-world uses (computer vision, warehouse automation, driverless trucks, drone delivery, and humanoid robots) and the risks of military AI, false recognition, and overestimating what these systems can do. A grounded, skeptical look at the AI hype.

Chapters

00:00 Intro: what this AI breakdown covers

00:40 Data, information, knowledge, and wisdom

01:31 Next-token prediction explained

02:24 Why machines don't understand meaning like humans

03:11 Training, loss functions, and token probabilities 04:08 Tokens, embeddings, and latent space

05:49 Why similar concepts cluster together

07:13 Numeric processing vs. human pattern recognition

08:29 How image models identify what's in a picture

09:49 Pretrained models, feature vectors, and similarity search

11:41 How humans and machines categorize new concepts

11:59 Military robots and autonomous systems

12:59 Ethics and safety before deploying robots in war

14:30 False recognition and friendly fire

15:11 Identifying friendly units with barcodes and QR codes

16:27 Automation in warehouses, trucks, and delivery

17:10 Driverless and cabless trucks

19:01 Drone delivery and the future of logistics

20:34 Artificial, not human, intelligence

21:16 Brain energy use vs. machine compute

22:00 How efficient the human brain really is

23:06 Is "AI" just a marketing term?

24:42 Fast retrieval vs. real reasoning

26:55 Do robots need the cloud?

27:57 Small language models for task-specific robots

29:28 Multimodal robots: vision, audio, and speech

30:19 Human-in-the-loop control

31:27 Robots for monitoring, industry, and heavy lifting

33:57 Where robots make sense now, and where they don't

34:30 Home chores and what kids lose if robots do everything

35:41 The future of autonomous defense

37:25 Final summary: AI as fast computation, not thinking

"It's just very powerful machines with very smart algorithms."

Follow No Fixed Lane for new episodes on tech, health, business, and life.