AI Design Technology: Safety First

TECHNOLOGIES ARE ADVANCING at an unprecedented rate, which is both exciting and terrifying. It’s tempting to think about safety in a dualistic sense — good and bad. The reality is more nuanced. We need to consider a spectrum of factors: Where the AI is deployed; how it’s deployed; the quality of the data it was trained on; and its impact on different groups of people. We have safety tests, crash tests, road rules, and driving tests to ensure that cars are used safely, and we can do the same with AI.

Yet, how do we prevent bad actors from abusing AI technology? Some AI platforms are taking a practical approach by focusing on three key areas: traceability, moderation, and content provenance. Traceability ensures that any generative AI technology can be traced back to a specific user, credit card, and address. Moderation involves controlling what users can generate within the platform. Content provenance helps determine whether the content is real or synthetic.

Emerging tools like classifiers, watermarking, metadata, and fingerprinting are becoming more common. These tools help identify whether AI generated a piece of content. Tools like Glaze and Nightshade, developed by the University of Chicago, can protect artists’ work and even create distortions in image generation. However, when exploring this technology, the makers rejected our request to test it because of our association with AI. This response is very dualistic and makes navigating the space more challenging.

AI transparency and accountability are still evolving, but some fundamental guidelines are emerging. The General Data Protection Regulation (GDPR) includes data protection, privacy, consent, and transparency provisions. The Organization for Economic Cooperation and Development (OECD) promotes trustworthy, transparent, explainable, accountable, and secure use of AI. The U.S. Government Accountability Office (GAO) outlines responsibilities and liabilities in AI systems. The EU Artificial Intelligence Act aims to regulate the development of AI systems in the EU.

Experts warn that AI’s increasing data hunger could incentivize unethical practices like violating data minimization principles or scraping personal data across the Internet. Data minimization means only collecting, processing, and storing the minimum necessary data. Explainable AI aims to make AI systems transparent and understandable to humans. This is vital for building trust and ensuring accountability.

Currently, this is uncommon. Perplexity AI is a tool that handles explainable AI fairly well. Perplexity, a Google alternative, is an AI chatbot-powered research and conversational search engine that answers queries using natural language predictive text and sources for its responses.

Slop: Image Spam

Slop refers to the Gen AI image equivalent of spam. These often are surreal images generated by AI. While some argue that the images have artistic value, they raise interesting philosophical questions about the nature of AI-generated content. For example, a recent Washington Post article showed a little boy painting a beach landscape and asking for likes. It’s a modern rendition of René Magritte’s famous work, La Condition Humaine, 1933 — a painting of a painting that blends into its background. The boy isn’t real, and the painting isn’t real. AI fabricated the entire scene, and the responses may be AI-generated.

AI isn’t just limited to images. AI-generated video content also is gaining traction, albeit with some hiccups. AI-generated videos can bring old photographs to life and even animate old paintings. But they can have too many legs or body distortions, as in early Generative Image AI. While some of these applications are fascinating, others are downright creepy.

What AI Means to Us

Let’s explore some AI tools for use in apparel embellishment. Ethical AI art-generation techniques like artist sampling involve using AI to analyze the style of a particular artist and generate new works that mimic that style. The artist is reimbursed, but the AI generates the actual image. Another example is Wand AI, which allows users to draw using an Apple Pencil and iPad; the AI quickly cleans up and refines the linework. This application preserves the original artist’s intention while streamlining the workflow.

AI Agents: The Future of Autonomous Workflows

AI agents are autonomous entities designed to perform specific tasks. They can range from simple chatbots to complex systems capable of managing workflows. Here are some examples available through °FAI and the Ink Kitchen:

• Meeting record agent: These are designed to transcribe and summarize meetings in real time. It can identify key points and action items and assign tasks to team members. This not only saves time but also ensures that everything important is noted.
• Sales Assist Agent: This can analyze customer input to provide personalized recommendations, improving customer satisfaction and boosting sales. This Agent can count colors from a PNG, recommend print methods based on factory capabilities, and be an expert in an apparel brand, making specific apparel recommendations for a project. This could be an onboarding agent for any salesperson outside of our industry or an assistant for a customer service rep.
• Content Creator Agent: This is a tool that can generate text, images, and even video. It can be used for marketing, social media, and content creation. While this Agent can save time and resources, it raises questions about originality and authenticity. It’s best to use it in tandem with human-created content.
• Ink Kitchen Agent: Designed to provide real-time information, it answers to queries relating to IK content. It can be deployed in customer service, education, and other fields and modified to contain any database of information.
• Research Agent: This tool can sift through vast amounts of data to find relevant information, making it valuable in research and development. It can help researchers stay up to date with the latest developments and suggest new areas of study. This could be applied to learning new dying processes, new application methods, new equipment, or pricing comparisons with the competition. (This is called AI Dynamic pricing in e-commerce and is rapidly rising in use).

The rapid pace of AI development brings both excitement and caution. It’s up to us to ensure that these technologies are developed and deployed responsibly. Watch °FAI on the Ink Kitchen YouTube channel to stay up to date on the fast-paced growth and use of AI and its impact on our industry. Stay curious, stay informed, and, most importantly, stay ethical.

*Human-written article with help from AI Agents built and deployed by the author.

Advertisement