107 | 👽 🤯 No Code. No Problem.

Brainyacts #107

In today’s Brainyacts we:

  1. do a walk-thru of an amazing new tool within OpenAi

  2. profile how AI detectors punish non-English speakers

  3. learn that when AI gets curious, it can outsmart a mouse?

  4. share a bunch of upcoming AI events

  5. see a comedian tell OpenAI it is not funny - and sue them

  6. coin “to Schwartz”- applying it to a South African lawyer

  7. learn how Hong Kong & China use AI in courts

  8. ponder where Japan is in the AI race

  9. learn Google and Mayo Clinic have partnered

  10. consider genAI for mid-size law firms looking to beat BigLaw

  11. hear Google fears deep fakes and is throttling its own AI

  12. question if ChatGPT can write AI regulation

Hi 👋 Brainyacts Readers. Today we gained a bunch of new subscribers. We are now over 2400!

Given the recent influx of new readers, I wanted to share some resources that some of you might be looking for or should know about.

  1. For newcomers to generative AI and ChatGPT, here is a link to my free email course. No Spam - Just an easy 5-day journey to get you up and running. Very practical and pragmatic.

  2. To read previous posts (all 106 of them), click here. If you are new to this, check out some of the early ones as they may be helpful. Just note much has changed in the world of generative AI but most things I covered will still be relevant.

Big announcement for all!

I am launching a self-paced 100% online (web and mobile) course. I call it the Brainyacts Generative AI Blueprint. To learn more, including seeing all contents and a preview video, click here.

Note prices will increase after August 1st - the official launch date.

And now onward into the newsletter . . .

👽 🤯 What is Code Interpreter and why you MUST try it

Too curious to read? Just watch my video below.

Imagine having a powerful tool at your disposal that can do complex math problems, work with large files, write and run code effortlessly, and surprise you with unexpected and amazing results. OpenAI's latest innovation, Code Interpreter for ChatGPT, brings this vision to life. You do not need to know how to code. This tool is designed to help everyone.

So, how does Code Interpreter work? Think of it as a super-powered assistant for artificial intelligence. By enabling the AI to interact directly with code, it becomes better at both math and language tasks. This enhanced capability ensures more accurate results and instills greater trust in the AI's output.

One of the most appealing aspects of Code Interpreter is its simplicity. You don't need to worry about coding intricacies because the AI does all the work for you. It's like having a skilled coder by your side, ready to execute any task. Whether it's working with data, analyzing it, or deriving insights, Code Interpreter simplifies the entire process. Uploading data, combining it, and cleaning it up are all effortless tasks for the AI.

It even has the ability to catch and rectify its own mistakes, further enhancing its reliability.

When it comes to data analysis, it shines with its remarkable intelligence. It can effortlessly navigate through datasets, uncovering valuable patterns and presenting them in a visually appealing manner through plots, tables, or charts.

Another remarkable use case of Code Interpreter is image processing. With its prowess, the AI can adapt and manipulate images in numerous ways. Whether you need to resize, crop, rotate, filter, or transform images, the AI has got you covered. Want to create captivating GIFs from images or videos? It can handle that too. Need to translate images into text or vice versa? No problem! It can even perform basic video editing tasks like applying slow zoom or fade effects. The possibilities are endless when it comes to harnessing the AI's image processing capabilities.

Additionally, if you do know how to code, Code Interpreter doubles as your personal code critic. It assists you in enhancing your code by identifying syntax errors, style issues, performance snags, security risks, and best coding practices. It provides valuable insights that can improve the quality and efficiency of your code. Moreover, it can generate documentation, comments, or tests for your code, further streamlining your development process.

In short, this is like a magic key that unlocks a vast universe of possibilities in data analysis, image manipulation, and code refinement. It empowers you to explore and analyze data effortlessly, transform and manipulate images seamlessly, and enhance your code effectively.

Now, witness the power of Code Interpreter firsthand. In the video below, I walk you through a simple yet captivating example. I uploaded a file and asked it a brainless question, expecting a mundane response. However, prepare to be amazed as the reply stunned me with its intelligence and creativity.

🌍🕵️‍♂️ The Language Conundrum: Non-English speakers keep getting flagged by AI detectors

While generative AI continues to be adopted by people, businesses, and governments and gain in acceptability generally, there is an unintended consequence surfacing in this landscape: a growing challenge for AI detection mechanisms to discern human-generated text from AI-produced content. The situation exacerbates with non-native English speakers, leading to a significant challenge for businesses and educational institutions globally.

These advanced AI detectors, touted for their high accuracy, are ironically flagging non-native English speakers' writings as AI-generated, undermining their efficiency. 

While AI detection tools are considered critical countermeasures to deter AI-enabled cheating, the issue is not so black and white. Businesses and academic institutions rely heavily on these detectors, especially in assessing essays, job applications, and scholarly work. However, the often alleged 99% accuracy by many of these detectors, appears to be misleading, particularly when the context involves non-native English speakers.

The discrimination stems from the way detectors discern human from AI-generated content. A core measure used by these programs is 'text perplexity'—the predictability of the next word in a sentence based on the preceding text. Generative AI models like ChatGPT are trained to produce low perplexity text, suggesting a smoother, more predictable language flow. However, non-native speakers, with their simpler and more predictable language patterns, often end up having their work flagged as AI-generated.

This flawed detection mechanism poses far-reaching implications:

  • In education, it potentially subjects non-native students to false accusations of cheating, impacting their academic career and psychological well-being.

  • For businesses, the language bias could unjustifiably dismiss valuable job applications.

  • Further, the problem trickles down to digital platforms where content flagged as AI-generated may be downgraded, marginalizing non-native English speakers' online presence.

An unconventional solution to overcome this lies not in enhancing AI detectors but in promoting an academic and corporate culture that encourages the creative and ethical use of generative AI. It calls for a shift from a battle against AI to collaboration with it. A radical approach would be to train AI models to mimic the linguistic diversity of non-native speakers, making them more inclusive.

Moreover, incorporating the principles of fairness, accountability, transparency, and ethics (FATE) into AI model design can help mitigate biases. Businesses and academia should prioritize efforts to understand and address the potential biases in AI and incorporate checks and balances that ensure diversity and inclusion.

Lastly, it might be beneficial to reconsider our reliance on automated AI detectors and invest more in human judgment (music to many humans’ ears). As AI detectors may not yet fully grasp the complexities and subtleties of human language, the discerning eye of a human evaluator still holds significant value.

⚖️ What might lawyers consider here?

Lawyers might find numerous legal issues arising from the language bias in AI detection mechanisms, with implications spanning discrimination laws, privacy issues, unfair competition, and intellectual property rights.

Here are some of the conventional and unconventional legal challenges I came up with - there are likely many more:

Conventional Legal Issues

1. Discrimination Laws: The most glaring issue relates to discrimination laws. If non-native English speakers are consistently discriminated against by AI detection systems, it could potentially lead to violations of equal opportunity employment laws and anti-discrimination legislation.

2. Privacy Laws: If an AI detection system is used to analyze written works without the author's consent, it might infringe on privacy laws. This is especially relevant if the AI system is used for educational purposes and if minors are involved. Or if all content, including emails, are automatically screened.

3. Consumer Protection: If AI detection tools are marketed as 99% accurate but consistently fail with non-native English speakers, there could be potential for claims under unfair competition laws. Misrepresentation of a product's capabilities can fall foul of consumer protection laws.

Unconventional Legal Issues

1. Linguistic Rights: Although not a traditionally recognized legal field, linguistic rights have increasingly gained attention in legal circles. Non-native speakers being incorrectly flagged as using AI could lead to a novel argument that their linguistic rights are being infringed upon.

2. Ethical AI and Algorithmic Accountability: These concepts, while relatively new, are gaining traction in the legal community. As the AI field advances, calls for laws requiring ethical AI and algorithmic accountability are growing. AI systems that unfairly flag non-native speakers could potentially violate these principles.

3. Intellectual Property: As AI use increases, unconventional IP issues may arise. If an AI detector tool flags content that is being submitted or used in an IP application or proceeding, what issues might this create? How can the human author prove they indeed write it and not AI? What burdens does this create?

Addressing these issues legally will require an innovative approach to lawmaking and enforcement, with lawyers at the forefront of shaping and interpreting new laws and principles. Thus, it's vital for legal practitioners to remain updated on the technological advancements and ethical implications of AI systems.

🚀🧩 Curious Replay: Helping AI learn in the “now”

A mouse outperformed an artificial intelligence (AI) agent in a simple exploratory task, sparking a groundbreaking study by interdisciplinary researchers from Stanford. The team introduced a red ball in both a physical enclosure with a mouse and a 3D virtual environment with an AI agent. While the mouse quickly interacted with the ball, the AI was oblivious to it.

This led the researchers to introduce "curious replay" to the AI's training - a method of prompting AI systems to reflect on and learn from novel and interesting encounters. The addition of curious replay not only enabled the AI agent to interact with the red ball faster, but also significantly enhanced its performance in a creative problem-solving game called Crafter.

The curious replay technique, an evolution of the existing "experience replay" method, prioritizes the reiteration of interesting experiences over repetitive scenarios. The innovation takes inspiration from studies on how the human brain, particularly the hippocampus, strengthens memories through "replay" during sleep.

The success of curious replay implies its vast potential in future AI research, with applications ranging from household robotics to personalized learning tools. This research will further inspire AI development from animal behavior, while also contributing to our understanding of animal behavior and neural processes.

In short, we continue to build the AI brain based on the animal brain. Given the animal world still relies on evolution and survival of the fittest/smartest . . . well, read into that any way you want. 🙂 

If you want to go deep, here is a link to the paper.

📆 🚀 Upcoming AI Events

News you can Use: 

Was this newsletter useful? Help me to improve!

With your feedback, I can improve the letter. Click on a link to vote:

Login or Subscribe to participate in polls.

DISCLAIMER: None of this is legal advice. This newsletter is strictly educational and is not legal advice or a solicitation to buy or sell any assets or to make any legal decisions. Please /be careful and do your own research.8