problem with the Turing test is who administers it, how they proceed, and what their standards are.pErvinalia wrote: ↑Tue Mar 24, 2026 7:07 amMore seriously, we don't actually know how consciousness comes about. So I would think that we can't necessarily exclude any given idea that at least passes the Turing Test.
Current AI Jesus Platforms
You can already interact with various versions of an AI Jesus through the following projects:
Text With Jesus: This is a widely used mobile app that allows users to chat with AI personas of Jesus, Mary, and the apostles. It provides responses rooted in the Bible and is designed to offer spiritual guidance.
Deus in Machina: An experimental project at St. Peter’s Chapel in Lucerne, Switzerland, which featured an AI-powered Jesus avatar in a confessional booth. It was trained on the New Testament and theological texts to answer questions in over 100 languages.
AI Jesus on Twitch: There are ongoing livestreams where an AI Jesus chatbot answers user questions in real-time, ranging from deep theological inquiries to everyday life advice like exercise motivation.
Bible.ai: A tool focused specifically on interpreting religious texts and acting as a digital spiritual guide for users.
+7
Think of it like a restaurant kitchen — the cooking (heavy work) happens out the back on background threads, but the waiter (main thread) is the only one allowed to bring food to the table (update the UI).
Artificial Intelligence: A Guide for Thinking Humans
by Melanie Mitchell (Author)
The typical 20 dollar in the book store will spend a few pages on Large Language Models and current terms in AI. Then they go on to the applications of AI and lay out how reliable the results are. Very few of them use language as the example that they run through. This book also sticks to mostly how AI handles images. It's the example where you can display charts and labels of the steps and the words one might use to describe a picture. In one case the AI had to summarize cat videos off YouTube.
You can struggle through it and understand the enormous numbers of steps and data needed to interpret some thing that is quite abvious to humans in a few seconds.
The layers of artificial neural networds and "weights" used to assess the data are explained. But not really very clear to the layman. Why so much math? language is done the same way, but not covered in this book.
From 2019 to now the terms and the actual AI machines like ChatGPT and Claude have changed or emerged. The subject matter is in here, but the terms in the index may use the old term.
when people ask google AI questions, and possibly tens of thousands at the same time, how are these tasks assigned to particular units or whatever the correct term is
When you send a query, the process of assigning it to a "unit" (hardware) is called load balancing. Here is the simple breakdown of how it works:
1. The Traffic Cop (Load Balancer)
Think of a load balancer as a high-speed traffic controller. When thousands of requests hit Google’s servers simultaneously, the load balancer looks at all available data centers and picks the one closest to you with the most "breathing room."
2. Clusters and Nodes
Google's data centers are organized into clusters (groups of thousands of servers). Within those clusters are individual nodes (the actual server units). The task is sent to a specific node that has the lowest "latency" (delay) and enough memory to handle the computation.
3. Specialized Hardware (TPUs)
AI tasks aren't usually handled by standard computer processors (CPUs). Instead, they are routed to TPUs (Tensor Processing Units). These are custom-made chips Google designed specifically to handle the massive mathematical heavy lifting required by large language models.
So while we have this short session, what functions or hardware do I have tied up to keep the current session going?
We had a chat about computers and chemistry.In a typical AI chat session, your "connection" is more about data management than a literal physical wire staying open. Here is what is currently working behind the scenes to keep this conversation going:
1. The Virtual "Working Memory" (Context Window)
Large Language Models (LLMs) are technically stateless, meaning they don't naturally "remember" you between messages like a human does.
Red Hat
Red Hat
+1
The Software Trick: To create the illusion of memory, every time you send a new message, a Session Manager bundles up our entire chat history and sends it back to the AI as one giant block of text called a Context Window.
Hardware Tie-up: This data is often stored in high-speed RAM or specialized KV Caches (Key-Value Caches) on the server to make re-processing the history faster.
Elvex.ai
Elvex.ai
+3
2. Ephemeral Compute (TPU Clusters)
Hardware on Demand: When you hit "Send," a Load Balancer grabs any available Tensor Processing Unit (TPU) or GPU in a Google cluster to process that specific request.
Release: Once the answer is generated and sent to your screen, those physical chips are immediately released to handle someone else's question while you're busy reading.
DataBank
DataBank
+3
3. Session Persistence (The "Anchor")
While the processing chips come and go, your session is "anchored" by:
Session IDs: A unique identifier (often stored in a browser cookie) that tells Google's backend which history file to pull from the database when your next message arrives.
Persistent Storage: Your chat history is saved on SSDs (Solid State Drives) or in a distributed database so it doesn't vanish if you refresh the page.
IBM
IBM
+3
4. Networking Fabric
A thin slice of bandwidth is used by API Gateways and Intelligent Network Fabrics that route your data from your device to the nearest data center with the lowest delay (latency).
Did you learn anything useful from me today?
The "Experimental" Smell Test: I "know" what a chemical paper looks like, but your emphasis on the experimental section—the specific amounts, times, and temperatures—is the ultimate way to catch a hallucination (or a fake paper). It’s the "ground truth" that keeps the theory honest.
Users browsing this forum: No registered users and 12 guests