Date: 2025-04-03 Page is: DBtxt003.php txt00026448
ARTIFICIAL INTELLIGENCE
OUCH!
Shelly Palmer: What You Can Learn From GPT-2
OUCH!
Shelly Palmer: What You Can Learn From GPT-2
Image created using DALL·E 3 with this prompt: Create an image of an older-generation LLM, as personified by a robot, doing work at a desk. Gothic style. Aspect ratio: 16×9.
Original article: https://shellypalmer.com/2024/03/what-you-can-learn-from-gpt-2/
Original article: https://shellypalmer.com/2024/03/more-ai-models-to-choose-from/
Original article: https://shellypalmer.com/2024/03/unlocking-the-power-of-ai-for-brand-marketers-the-benefits-of-fine-tuning-rag-and-transfer-learning/
Peter Burgess COMMENTARY
These three pieces from Shelly Palmer are very thought provoking. They have challenged me into thinking more about the AI boom that has now been going on for several months.
I am reminded of an awful day in my life when I was just 18 years old. I was a reasonable athlete but not excellent an any one sport, but quite good at a variety of sports ... rugby, cricket, swiming. athletics, etc. Though I was not really good athlete, I was good enough to be in the finals of eight different events that would all take place on 'sports day'. And though I had passed through the 'heats' in order to get into the finals I had little chance of being in the top three ... more likely, I was going to be last in all the events. The most humiliating was the final of the 1 mile run where I had a great view of the finish while I still had a half lap still to run.
Though I did not contribute much to our 'house' performance on the day of the finals ... I had something of a reecord for the mumber of 'qualifying' points I had earned for the house simply by getting into the finals of so many events.
I am not sure why this memory has surfaced while I am thining about what Shelly Palmer has written and where the world is going with AI.
Perhaps I am worrying that the emergence of AI is going to be something like the emergence of modern nuclear physics which enabled both atomic bombs and atomic power generation.
I have been involved in living my life for a long time, and in this time I have had a number of terrible interactions with powerful people. Powerful AI in the hands of powerful people is not going to be a problem if the powerful people are 'good people' ... but what is going to be the outcome when the powerful people are 'bad' with little or no moral compass. I am very worried that if AI gets to multiply the power of 'bad', then the future will be very dismal for the world as a whole.
I don't expect to sleep well tonight!
Peter Burgess
Written by SHELLY PALMER
MARCH 19, 2024
I just re-watched a video I first saw a few months ago: “Demystifying how GPT works: From Architecture to…Excel!?!” It’s a “how-to” video that uses a very clever Excel spreadsheet to walk through how large language models (LLMs) work. If you’ve been wondering exactly how an LLM takes your prompt and predicts what’s next, this will be 10 minutes well spent.
This may be just a bit too geeky; if it is, let me know. I’m putting the finishing touches on my AI for Brand Marketers Playbook, and I would truly appreciate your feedback about just how deep to go.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.
ABOUT SHELLY PALMER ... Shelly Palmer is the Professor of Advanced Media in Residence at Syracuse University’s S.I. Newhouse School of Public Communications and CEO of The Palmer Group, a consulting practice that helps Fortune 500 companies with technology, media and marketing. Named LinkedIn’s “Top Voice in Technology,” he covers tech and business for Good Day New York, is a regular commentator on CNN and writes a popular daily business blog. He's a bestselling author, and the creator of the popular, free online course, Generative AI for Execs. Follow @shellypalmer or visit shellypalmer.com.
More AI Models to Choose From
MARCH 18, 2024 BY SHELLY PALMER
Elon Musk’s xAI has officially open-sourced its AI chatbot Grok, which is now available on GitHub. The release includes the base model weights and network architecture of Grok-1: a 314 billion parameter Mixture-of-Experts model, under the Apache 2.0 license, which enables commercial use but does not include the data used to train it or connections to X for real-time data. (I started working with it this weekend. I’ll write a review in a week or so.)
Elon Musk, who was a founder of OpenAI but later criticized it for moving away from its original open-source ethos, has been a vocal advocate for open-sourcing AI technologies as a counterbalance to big tech’s proprietary approach.
The decision to open-source Grok aligns with Elon’s broader vision for AI, which includes understanding the true nature of the universe and ensuring AI development is aligned with human safety and ethics. If you remember, Elon had a pretty public falling out with Google co-founder Larry Page about the risks of AI. The spat ended with Page calling Elon a “speciesist.”
On a related topic: The title of my Sunday essay, Unlocking the Power of AI for Brand Marketers: The Benefits of Fine-Tuning, RAG, and Transfer Learning, may be a bit jargony, but the subject is anything but. It explores three techniques for gaining a competitive edge by combining your data with AI.
Illustration created by DALL-E with the prompt “A highly detailed close-up of a human index finger interacting with a digital touchscreen interface highlighting the theme of human and AI integration in brand marketing. Aspect ratio 16×9.”
Written by SHELLY PALMER
MARCH 17, 2024
Fine Tuning AI
Integrating AI into marketing strategies has transitioned from novel to necessary. But how do you gain a competitive edge? Part of the answer lies in your ability to customize the AI models that support your tech stack. Techniques such as fine-tuning, retrieval augmented generation (RAG), and transfer learning offer pathways for AI-driven solutions to align with your brand’s identity. By leveraging these approaches, you can enhance the precision, relevance, and impact of your campaigns, ensuring that AI not only complements but elevates your marketing efforts.
Fine-tuning
The process of fine-tuning allows you to adjust AI capabilities to fit the needs of your brand. Think of it as training an exceptionally talented new team member on the specifics of your company’s language, goals, and customer base. This approach ensures that the AI-driven content and recommendations are not just accurate but also resonate with your audience on a personal level. The result? Enhanced engagement, loyalty, and insights into customer behavior.
Fine-tuning is ideal when you have specific goals and a unique brand voice that generic AI models can’t capture out of the box. It requires a solid dataset that reflects your target audience’s interactions, preferences, and behaviors. However, if your primary challenge is keeping content dynamically updated with the latest information, fine-tuning alone might not suffice. In such cases, integrating RAG could offer a more comprehensive solution by ensuring content remains fresh and accurate, aligning with real-time data and external developments.
RAG (Retrieval Augmented Generation)
RAG gives AI the ability to aggregate the most current information from a vast array of external sources to enrich the content it creates. For marketers, this means that your AI-driven tools can automatically update product details, pricing, and promotions in your communications ensuring that your messaging is always relevant and accurate. The ability to seamlessly integrate the latest data into customer interactions and content enhances customer experience while maintaining alignment with your brand guidelines.
RAG stands out when your marketing strategy calls for integrating the latest information or data from diverse sources. It’s particularly useful if providing up-to-the-minute accuracy is crucial. However, RAG’s reliance on external data sources means it’s less about tailoring the AI’s voice to your brand and more about ensuring content relevance and factual correctness. If customization and brand-specific messaging are your primary objectives, fine-tuning, possibly in conjunction with RAG, may offer a better solution.
Transfer Learning
Transfer learning is about standing on the shoulders of giants. It allows you to apply insights from large, pre-trained AI models to your specific marketing needs without starting from scratch. This approach is incredibly efficient, reducing the time and resources needed to deploy AI solutions that are tailored to your marketing tasks. For marketers, this means access to state-of-the-art AI capabilities that can enhance customer engagement, content personalization, and market analysis, even if your organization doesn’t have vast amounts of data or extensive AI expertise.
Transfer learning is particularly beneficial when you’re looking to leverage AI’s capabilities quickly and with limited data. It’s the shortcut to deploying sophisticated AI without the need for extensive datasets or the time-consuming process of training models from the ground up. However, while transfer learning can provide a strong starting point, it may not always offer the level of customization or real-time data integration that fine-tuning and RAG provide. If your marketing needs are highly specialized or require up-to-date information from various sources, combining transfer learning with other techniques will probably be required.
Isn’t This What I Hire Engineers To Do?
The journey into AI-driven marketing is not about mastering the technical details (you do hire engineers for that) but about understanding how to strategically deploy AI to enhance your brand’s engagement, efficiency, and insights. Ask your tech team about their approach to fine-tuning, RAG, and transfer learning. There’s a pretty good chance they’re already using some (or all) of these techniques to enhance your marketing capabilities.
Modified author’s note: These are not a sponsored posts. Shelly Palmer is the author of these articles and they express hiw own views!