Hi folks, welcome back to another issue of AI Growth Pad.
In today issue, we add a section to bring you awareness about potential problems with AI and what can be done to mitigate it. We still a firm believer of AI potential to positively impact humanity. But it’s critical to be aware of how to protect yourself from its pitfalls or malicious uses
🛟 LLM is susceptible to supply chain attack
🛠️ Top 3 AI tools to try
⚡️ChatGPT code interpreter is game changing
LLM is susceptible to supply chain attack
Software supply chain attack was how thousand of corporate and US government networks were hacked in 2020 by Russian hackers
This came to be known as the SolarWinds attack.
Now researchers have shown that LLM, the machine learning model behind ChatGP, is also susceptible supply chain attack.
They show how one can surgically modify an open-source model to make it spread misinformation on a specific task but keep the same performance for other tasks.
More damagingly, they distributed on Hugging Face, the leading hub for open-source machine learning models undetected
Read more about the research here
Top 3 AI tools to supercharge your productivity
Since last week issue, folks have asked how we choose which tools to feature in our newsletter. Here are our criteria
Leverage AI: The tool has to use AI to some degree. Duh!
Ridiculously useful: The tool can 10x your productivity.
Mature product / company: It’s a bummer if your favorite tool closes shop next month. So we only consider companies that are either more than a few years old or well funded.
Framer
Framer is a no code website builder that can generate beautiful website from a simple prompt. Unlike others, Framer has been around for a while and has a robust visual editor so you can further edit your AI-generated site
I used Framer to generate a landing page for an idea I have and result is quite impressive. You can check it out here.
ChatGPT code interpreter
Originally announced in March along with ChatGPT plugins, code interpreter lets ChatGPT users upload data and it then automatically write code to perform data analysis. However, the potential of code interpreter goes much deeper as we see in the next section
Wondercraft.ai
Wondercraft enbles anyone to generate professional-quality podcasts using state-of-the-art LLMs and Text-to-Speech technology. Imagine turning blog posts, articles, and essays into captivating audio content in seconds.
ChatGPT code interpreter is game changing
Last week, OpenAI announced code interpreter is available for all ChatGPT plus subscribers. People are already proclaiming this to be the biggest update since ChatGPT launched.
To understand why, it’s helpful to know what code interpreter is
A environment running Python and over 330 libraries like pandas (data analysis), matplotlib (plotting), Scikit-learn, Pytorch and Tensorflow (Machine learning)
It allows users to upload up to 100MB including .csv, .xls, .png, .jpeg, .mov, .mp3, .pdf and .zip. It can automatically unzip files so you can upload more than 100MB in theory
The environment is sandboxed to a specific chat session. That means it does NOT have connection to the Internet nor can it access other chat sessions
Why is this a big deal?
Because ChatGPT can write code, execute it, observe the result and can correct itself all from a single prompt.
This opens up endless possibilities because code can be arranged in innumerable ways. Let’s observe what people have achieved with code interpreter to have a better grasp of its capability
Build and train a machine learning model https://twitter.com/swyx/status/1677575532597383169
Draw a meme https://twitter.com/swyx/status/1677567255377494016
Build a 3D visualization of Billboard Hot 100 https://emollick.github.io/3Dmusic/
Face detection
https://twitter.com/shacrw_/status/1677774686648713216
I’m learning how to apply this to practical business problems and will be writing a series about it. Lemme know if you’d like to get the first draft as soon as it’s available by answering the poll below
That’s all for today folks. If this was helpful to you, please share to your network and let us know how we can improve in future issues
A
About Trung Vu
Trung Vu, a former software engineer, founded Hoss in 2019 to enhance developer experiences, swiftly attracting Silicon Valley backers and a $1.6 million seed round. In 2021, his venture was acquired by Niantic Labs, of Pokemon Go fame, to bolster their Lightship platform.
Post-acquisition, Trung leads engineering teams at Niantic and invests in promising AI startups. An AI enthusiast even before ChatGPT's rise, he equates its potential to electricity. Through AI Growth Pad, his education platform, Trung teaches entrepreneurs to leverage AI for growth, embodying his commitment to ethical, transformative technology.
