مقایسه ۲ غول هوش مصنوعی؛ DeepSeek مقابل ChatGPT

https://www.tasnimnews.com/fa/news/1403/11/08/3245970/%D9%85%D9%82%D8%A7%DB%8C%D8%B3%D9%87-2-%D8%BA%D9%88%D9%84-%D9%87%D9%88%D8%B4-%D9%85%D8%B5%D9%86%D9%88%D8%B9%DB%8C-deepseek-%D9%85%D9%82%D8%A7%D8%A8%D9%84-chatgpt

On Polytopes

https://www.lesswrong.com/posts/GdCCiWQWnQCWq9wBE/on-polytopes

A Spline Theory of Deep Learning

https://proceedings.mlr.press/v80/balestriero18b.html

IT Worker Vandalism

https://techxplore.com/news/2025-01-worker-british-museum.html

Poor Performance of Devin AI

The "First AI Software Engineer" Is Bungling the Vast Majority of Tasks It's Asked to Do

https://futurism.com/first-ai-software-engineer-devin-bungling-tasks

AI in Cosmology

https://www.thebrighterside.news/post/new-ai-system-precisely-models-the-universes-evolution-after-the-big-bang/

Standard Work-Breakdown Structure for AI/ML - tidbit

There are common activities that take place during AI/ML development before it undergoes the process of turning it into a product and deploying it.  Those generic activities could be considered as elements of a standardized WBS for AI/ML development.  They are:

• Task 1 – Understanding the Problem

‒        Problem Definition

• Task 2 – Understanding the Data

‒        Data Acquisition

‒        Data Processing

‒        Data Cleaning

‒        Data Exploration

• Task 3 – Model Development (linear or non-linear)

‒        Model Triage

‒        Feature Vector Engineering

‒        Model Selection

‒        Model Training and Testing

• Task 4 – Model Deployment

‒        Develop Documentation

‒        Define Development stories and Narrative

‒        Development Handoff

‒        Support Development

‒        Support Deployment

‒        In Situ Evaluation and Iteration

‒        Long Term Support

Not all of these tasks and subtasks occur all the time in each and every development stream, but their commonality makes it possible to use them as templates for project management and tracking.  

These tasks and subtasks could be imported into such project management tools as Microsoft Project to reduce the cost of project initiation and setup.

In an analogous manner, these tasks and subtasks could be imported and relabeled as Epic/Feature/User Story in such tools as Jira, Jira Align, Azure DevOps to standardize the development process for AI/ML projects.

Furthermore, taking a baby-step towards the so-called "Trustworthy AI", one could use such things as GUIDs to tie a project to its WBS elements and then to tie the WBS elements to the development artifacts for each task & subtask and package and submit them to Blockchain.  This creates reliable curation of the AI/ML development artifacts and deliverables possible. 

AI Pollution

We read:

"AI is supposed to make things quicker and easier, right? But while it's true that AI tools can automate or streamline some of the most tedious tasks, it appears that the tech, or its results, are making some tasks more difficult."

https://www.creativebloq.com/ai/ai-art/designers-say-ai-is-making-stock-image-sites-unusable

Killed by LLM

From a reddit post @  https://www.reddit.com/r/LocalLLaMA/comments/1hs6ftc/killed_by_llm_i_collected_data_on_ai_benchmarks/

Killed by LLM – Collected Data on AI Benchmarks

For my year-end I collected data on how quickly AI benchmarks are becoming obsolete.

It's interesting to look back:

2023: GPT-4 was truely something new

• It didn't just beat SOTA scores, it completely saturated benchmarks

• It was the first time humanity created something that can beat the turing test

• It created a clear "before/after" divide

2024: Others caught up, progress in fits and spurts

• O1/O3 used test-time compute to saturate math and reasoning benchmarks

• Sonnet 3.5/ 4o incremented some benchmarks into saturation, and pushed new visual evals into saturation

• Llama 3/ Qwen 2.5 brought Open Weight models to be competitive across the board

Today: We need better benchmarks

• I'm amazed seeing tasks I didn't think we'd solve until 2030 become obsolete, and yet we still can't trust a model to do the same tasks as a junior

• It's clear our benchmarks aren't yet measuring real-world reliability, I hope we have as much progress in benchmarks as we do models in 2025.

Let me know what you think!

Code + data (if you'd like to contribute): https://github.com/R0bk/killedbyllm
Interactive view: https://r0bk.github.io/killedbyllm/

DeepSeek-R1 and exploring DeepSeek-R1-Distill-Llama-8B

From Simon's Weblog

https://simonwillison.net/2025/Jan/20/deepseek-r1/

News of AI from China

https://arstechnica.com/ai/2025/01/china-is-catching-up-with-americas-best-reasoning-ai-models/

AI Hallucinations

AI hallucinations can’t be stopped — but these techniques can limit their damage:

https://www.nature.com/articles/d41586-025-00068-5?utm_source=fark&utm_medium=website&utm_content=link&ICID=ref_fark

Electricity & Water Demands of AI

https://www.theguardian.com/technology/2024/sep/15/data-center-gas-emissions-tech

U.S. Tries to Govern AI’s Global Spread

From the Carnegie Endowment for International Peace:

With Its Latest Rule, the U.S. Tries to Govern AI’s Global Spread | Carnegie Endowment for International Peace

Testing HTML Page for jQuery - tidbit

 <!DOCTYPE html>

<html>

<head>

    <meta charset="utf-8">

    <title>jQuery Document Ready Demo</title>

    <script src="https://code.jquery.com/jquery-3.2.1.min.js"></script>

    <script type="text/javascript">

        $(document).ready(function(){

            $("p").text("Hello World!");

alert("Test: Inside the ready function.");

        });

    </script>

</head>

<body>

    <p>Not loaded yet.</p>

</body>

</html> 

Accessing the Location of a File in PCF - tidbit

Adding the file path in application properties does not help to find that file. 

Instead, putting file under src/main/resources and then using the following code helped to find file in PCF: File files = new File("./BOOT-INF/classes/file.csv").

The Symbolism of Guernica

 

Enterprise Programming vs. Startup Programming

 

Data Privacy

 

Resolving Chromecast-Hoopla Inter-operability Problem

I am describing below a Chromecast-Hoopla inter-operability problem & its resolution.

Description:

I noticed the problem first in October of 2024.  I would cast to Hoopla via Chromecast from my Android phone.  The TV screen showed the movie being loaded (the circling Wait icon) and then it would stop and would only display the blue Hoopla screen.

I verified the same behavior with Apple iPhone.

The specifications were:

• Device brand and model: iPhone 14
• Operating system, found in device settings: iOS 17.61
• hoopla app version: 4.60
• Chromecast model: firmware – 1.36.157768

and

• Device brand and model: Galaxy S8
• Operating system, found in device settings: Android 9.0 
• hoopla app version: 4.77.1
• Chromecast model: firmware – 1.36.157768

and

• TV: Sanyo
• Model No.: DP32242 HDMI
• Manufacturing Date: SEPTEMBER 2012

All devices were on a home wireless network.

Resolution:

I could not update the firmware on the Chromecast since there was no option to do so using Google Home app. on the Android phone.

I updated the firmware of the TV set from the Web (it was on the wireless network) via its remote-control options.

The Chromecast-Hoopla inter-operability problem was resolved.