Read Time: 2 minutes
I just wrapped up my last class on AI Marketing for 2024 at the University of New Hampshire, and I am always inspired by what attendees are working on.
Marketers from vastly different organizations, including Universities, Fortune 100 companies, and local SMBs, are all working to improve their use of AI to get more done in less time.
In class, we discuss building processes and pipelines (content, image, video, scraping).
But the core of that is taking the time to improve your prompting. No matter what model you use or what you are trying to accomplish, block time on your calendar to work on better prompting.
Be specific and prescriptive in prompts
Assign a role/persona to the AI (ie email copywriting expert)
Provide context (geography, time period, industry etc.)
Set boundaries and constraints
Use follow-up prompts to refine outputs
QA results by asking for sources and explanations
Google Gemini is the LLM I use the least, but here are the data points that are starting to change my mind:
NotebookLM ā You can upload text, notes, etc, to build your own personal RAG without coding. I tend to use Gemini when I need that larger context window, but how they connect it to audio is what makes it unique. >>> Watch more on how folks are actually using NotebookLM
Gemini Live ā Itās now free. You can talk 4x the speed you type. If youāre an Android user, itās worth seeing if you like it better than ChatGPT.
185 Case Studies of actual companies using Gemini.
Glasp is the new web clipper. I loved Evernote, but it died a slow death. I wouldnāt recommend adopting Glasp yet but the free version is worth checking out if you have a YouTube problem like me or read on Kindle and/or use Notion.
Opensource LLMs like Llama; and what a pissed-off and motivated Mark Zuckerberg will do.
A quick summary of what Zuckerbergās up to lately:
John Cena and Judi Dench are voice options. Celebrity will be an interesting driver of AI adoption.
Mark thinks smart glasses will replace phones.
Llama released a free vision-capable 3.2 version