The "Myth-Busting" Edition
Marco Andre
Top Voice 2024 | 50% more output in 50% less time. I train medium and large organizations to get ahead with AI | Global Keynote Speaker | Marketing & AI Executive
These are the top 2 questions I get asked on AI.
In today's edition I try to myth-bust them - "Who you gonna call"?
1. "If I put something in CHAT-GPT, is my company or private information immediately used to train the model - i.e. becomes public?"
- The answer is NO, ChatGPT does not automatically learn or get trained on user inputs during conversations. However, OpenAI does use some conversations with ChatGPT to help improve and fine-tune future versions of the model
- This process of incorporating user feedback happens offline, separate from any individual conversation, when OpenAI retrains the model on new data.
- Any user can optionally choose to allow OpenAI to use their conversations for potential model improvements, but this is not automatic and is something that can be changed in settings.
2. "AI ONLY works with a solid data foundation"
- Having a solid data foundation means having the right infrastructure, processes, in place to ensure high-quality, easily accessible data. In others words, your data is labeled and organised.
- This is YES essential for applied AI (think a Netflix or Amazon recommendation engine ) and yes, parts of Gen AI such as Custom LLM's (changing an existing model).
领英推荐
- If you want content to be based on your data (for example: your customers or products) you need that data foundation to be right. As an example, when you are trying to generate something out of something e.g. a doc of a series of docs, an Excel out of a series of data points.
- But in cases in which you only need to generate something from scratch - think tools like GPT and DALL-E for content like an image - you do not need to have your data properly labelled or organised. This is useful for instance when you want to do something more creative, not bound by previous iterations. Thank you Ben Torben-Nielsen, PhD, MBA for helping me with this one).
I understand why some of these myths exist. And some of them might have been true at some point, specially as we are at the early stages of consumer facing Generative AI and tech is evolving rapidly.
I welcome any fellow myth-busters that have a different opinion and can help shed some light.
Because we need to keep asking questions.
See you next week.
Hey, I’m Marco. I write about AI, Leadership and #ImperfectStories .
If you like what you just read, connect, follow me Marco Andre.
4x Founder | Generalist | Goal - Inspire 1M everyday people to start their biz | Always building… having the most fun.
5 个月AI job stealing a big one!?Looking forward to hearing your take on it.
Crescent CEO. Follow me to be part of our journey building a $10B+ company.
5 个月This is great Marco. Great clarity given!
Global Commercial Leader in Pharma / Driving Innovation & Commercial Excellence in Healthcare / Building Bridges
5 个月Thanks, very helpful
Business & Data Analysis | Business Intelligence | Market Research | Python (PCAP certified) | Power BI (Microsoft Certified) | Process optimization
5 个月I would also add myth busting that AI can actually think and reason. It cannot. By definition and model constitution this is an imitation and it “guesses” next word based on probability. It does not understand the meaning but sometimes imitation is very helpful, like in call centres or basic reports writing. Painful topic for me, serious people talking that AI will replace dietitians, so I’m working with a portal for dietetics and nutrition professionals on a series of articles explaining how to collaborate with AI and remain irreplaceable. P.S. I heard about Chat GPT and Google AI (long before Gemini’ “passing” Turing test. It didn’t if you read details and how the test was constructed. But I love Washington Post for creating a buzz.
Building ? performance teams that deliver true digital transformation
5 个月On number 1 though, my friend Marco Andre ??... let's not be as naive as to think they won't use it because they will. How - we may not be 100% sure. But that's the currency of the exchange, data.