AI Tech

Site Name

AI Tech

Description

Technical AI discussion. Questions, answers and articles about creating and using sophisticated cutting-edge AI models.

Topics

Topics covered would include:

• Latest models in existence and their capabilities
  • Examples: GPT models, DALL-E, Llama, ...
• Suitability of different models for specific tasks
• AI model API usage
• Prompt engineering
• Model deployment
• Design of software systems incorporating AI models, with a focus on how the model fits in
• "Boosting" simple ML models off of output from sophisticated models
• Downloading open source models to run them locally
• Training sophisticated models
• Prompt tweaking to improve the quality of model responses
• Optimizing financial costs of using models
• Probability and statistics as they pertain to the use and creation of sophisticated AI models
• Wrappers, clients and software tools that improve the usability of AI models
• Practical questions about maintaining security and privacy when using 3rd party models
  • Examples: Sandboxing model activity, clarification of service ToSes, questions about real past behavior of model service providers (in case they have not followed their ToS)
• Practical questions about AI ethics and safeguards, such as how to implement a specific principle via prompt engineering or software
• Recommendations about AI are on topic
  • Examples: Specific AI models, AI-related software, AI related books, articles and other resources, hardware (including consumer components as well as cloud architecture) for running heavy AI workloads, large training datasets

Note that there are no special restrictions regarding AI generated content on AI tech. Posts are subject to general Codidact rules on AI content, but AI tech is intended to allow whatever Codidact allows. The primary metric for moderating posts on AI tech is the quality and accuracy of the post rather than its source.

Exclusions

These topics or types of posts would be out of scope:

• Simple models and ML techniques, like linear regression
  • As a rule of thumb, if the ML topic was well understood prior to 2015, it is probably not a good fit for the site
  • Another good rule of thumb is that if training a model is trivial on consumer hardware, it is probably not a good fit
• Posts not about AI models
  • As a rule of thumb, if your post doesn't mention AI models, it is probably a poor fit
  • General posts about: Software engineering, probability, statistics, data science, data analysis
• Speculative, sensationalistic, non-technical or overly broad "big picture" questions
  • Examples: AI ethics (except for technical discussion), social implications, "responsible AI", philosophy of conciousness
  • These are important topics, but the public dialogue is currently dominated by non-technical and ill-informed opinions. They are not a good fit for a site intended for expert discussion of technical AI matters. We direct users to sites focused on philosophy, ethics, politics and laws.
  • Example questions which are off-topic:
    • Does the use of AI violate human rights?
    • Should AI have rights?
    • Is the use of AI a kind of slavery?
    • Is it mean to hurt the AI's feeling?
    • Is AI actually conscious or just creating the illusion thereof?
    • Is it ethical to use AI for weapons development, warfare, finance, housing or criminal justice?
    • Will AI put X profession out of a job?

Special Features

Would this community want additional categories, new post types, changes to site settings, etc? Codidact has a lot of configuration available, so if there's something you want to be able to do, please share it.

Currently none. If you have an idea, please create a post in https://proposals.codidact.com/categories/68

Overlaps

There are substantial overlaps with a number of areas. However, it's important to keep in mind that:

• Lately, "AI fatigue" has grown as many people are tired of hearing about AI in the news and other channels. So in many cases of theoretical overlap, there is probably no problem, because the overlapping site would rather exclude AI discussion anyway even if it seems to fit under their general subject.
• The overlap is usually in the auxiliary aspects of developing and using AI, while discussion of the AI itself (which is the heart of the matter) tends to be off-topic. So discussing the same thing in overlapping sites must necessarily be round-about, whereas on AI tech the focus is on the AI specifically.

Software Development

• Prompt engineering is off-topic on SD: https://software.codidact.com/posts/288842
• Questions about designing AI systems tend to be too broad, example: https://software.codidact.com/posts/289009

Generally, for discussion about auxiliary programming related to AI, such as writing API clients, the logic is similar to Game Development: If the questions is about both programming and AI, it is technically suitable on both AI tech and SD, but it is more suitable on AI tech and is more likely to get useful responses from an audience of people interested in the subject.

Power Users

It could be argued that the use of AI falls under the umbrella of "using software", but that is akin to saying C questions should go on Power Users because they're about "using the C compiler". The scope of Power Users is quite broad and not all AI discussion is necessarily a good fit. Therefore, AI-related questions should go in AI tech rather than Power Users.

Linux

Much AI-related software runs on Linux. However, the specific shell commands used to execute the software are rarely a critical factor in accomplishing the tasks that AI is used for.

(Hypothetical) Politics

Discussions about various social implications of using AI and how these uses should be governed tend to focus not on the technical details, but on the political aspects of AI. Therefore they are a better fit for a politics-focused section in the vein of politics.stackexchange.com.

(Hypothetical) Legal

Discussions about what is or should be legal with respect to AI usage is more concerned with the legal, social and ethical implications rather than technical details of how the AI works. These are a better fit for a law-focused section in the vein of law.stackexchange.com.

(Hypothetical) Philosophy

Discussions about whether the AI is truly conscious, and what philosophical ideas apply to it, are more concerned with ethical and philosophical implications than the technical details of how the AI works. At this point, AI tech discusses "AI" as an inanimate mechanism no different than any other software tool, with no special philosophical consideration is needed. Philosophical discussions about AI are a better fit for a philosophy-focused section in the vein philosophy.stackexchange.com.

(Hypothetical) Statistics and data analysis

While AI models are typically implemented using a broad mathematical foundation in probability and statistics, we view these as implementation details rather than the point of the AI. When such matters directly influence the use of an AI, they are on topic for AI tech. However, general questions about statistics or data analysis should be on their own section, in the vein of stats.stackexchange.com.

posted over 1 year ago

1y ago by ArtOfCode‭

matthewsnyder‭

141 82 73