Mission
The ZiptieAI mission is to give you
- 1 Recognition of what works
- 2 And what doesnt work
- 3 Hands-on demos (QS)
- 4 Intrinsic understanding of core concepts
- 5 Deep dive (NNs, Models, Agents, Projects)
- 6 Anticipation of the future
1 Recognition of what works
Recogition and acknowledgement of reality is the most important aspect of a successful AI application.
Palantir’s initial projects were for military applications without AI (25 years ago). They were very successful. The applications were basically agentic apps (without AI). The apps were designed to help decision makers. Not to replace mankind. AI later became a very valuable helpful assistant. Palantir’s missions are perfect for AI-assisted agents. Benefits outweigh the negatives. Errors can be tolerated. And a human always makes the final decision. Its no accident that Palantir is talking straight about many aspects of AI. Engineers wont go away, their expertise is required in the AI age more than ever. Palantir’s business does not demand the myth of AI intelligence. Palantir says that AI will not replace people, it will empower them.
A successful mission for AI
2 And what doesnt work
Falling for AI hype (as a product dev or consumer) will cost you dearly. “No more programmers” (Jensen Huang), “no more workers” (Musk, the colonizer of Mars), “we must pause AI development because its not safe” (Anthropic CEO), “AI already has emotions” (Nobel and Turing prize winner Jeffrey Hinton) etc etc. The general public has no idea how AI really works. Fertile ground for con artists.
If AI was truly intelligent then AI integration would just be connecting AI and letting it rip. But AI is not intelligent. AI is based 100% on deterministic binary algorithms that compute statistical probabilities. AI is an addon, a helpful assistant. The most reliable AI systems use the smallest AI component that can provide the required functionality.
AI is a security risk. The core NN in AI has what are often referred to as “hidden layers” that perform the massive brute-force number crunching to generate “intelligent” output. It’s impossible (even for AI) to determine how a NN has been programmed by analyzing the parameter settings in these layers.
An unsuccessful mission for AI
3 Hands-on demos (QS)
You can only learn how AI works by doing code demos (or at least studying the code). Chatbots are programmed to inundate you with hype and AI lingo. GPT often remarks about my habit of wanting to describe AI from a mechanistic viewpoint. My response has always been “of course, AI is mechanistic (binary computation).
A simple NN demo of the core of AI
4 Intrinsic understanding of core concepts
AI is first and foremost a man <> machine interface. When you chat the interface is via tokens (words). Human eyesight and intelligence converts these tokens into thoughts. An amazing little-understood process. Tokens are converted to numbers before input to an LLM. An LLM see nothing, has no thoughts, it only crunches numbers. NNs are always passive; CPU-based agents are always in control. An LLM NN inputs data and spits out a (100% deterministic) result. Thats it.
Human (left), agent (center), and NN (right)
5 Deep dive (NNs, Models, Agents, Projects)
Without AI tools (chatbots) I never could have accomplished 5% of what I have done on the ZiptieAI project. But AI chatbots only say what they are programmed to say (sometimes that’s pure AI lingo and hype). I went down so many rabbit holes when studying AI. The ONLY way to understand AI is by doing examples. Code does not lie.
Code for NN definition (left) and training (right)
6 Anticipation of the future
AI has big a future even after the pre-IPO hype is over. Those who understand how to integrate AI into their own workflows will benefit the most. ZiptieAI has a mission to help myself (and others) stay ahead of the AI curve. In the near future, AI integration skills will be required in the digital infrastructure of all business segments (just like the PC 40 years ago). This work will required a massive number of tech workers who, empowered by AI tools, will help integrate AI as a helpful assistant into all areas of life.
The helpful assistant of the future
26.0615 (v1 26.0611)