Report from conference AI Restart
What was interesting for me?!
The Czech Republic isn't a big country - so we often meet at conferences and other actions. Today I want to share with you some ideas from yesterday's conference called AI Restart which kick me back in the chair to be more involved in AI. The first part is for all - second will be only for friends of advanced SEO.
So there are some tips from AI restart that I discovered yesterday and they should be interesting for you too. So let's begin what's new in the world of SEO after AI Restart 2024.
Keyword Analysis in SEO
There was a great presentation about implementing LLM in the keyword analysis process which should be boring. I love to see others thinking about the same as I think because they often come up with different solutions and that should be inspiring.
Lukáš's flow of keywords analysis was about using NLP to extract n-grams that are statistically significant for that dataset and after that, they make classification groups with AI. These groups they uploaded into the KeywordR package and after that, they were classified for keyword analysis.
The author said that they should save 80 - 90% of time on classification and from 10 - 20 hours in which they work on keyword analysis they are on 1 - 3 hours. And because they use multiple prompts they have different times to check outputs and update them.
It isn't perfect because, in an ideal world, I'd probably like to have a job without the 1-3 hours they do now, but it's an excellent progression. The main point is that they should focus a lot more on interpretation and to bring outcomes to clients…
… and that's important I think. If AI or LLMs bring something to the table is that we should do less “clicking” work and be much deeper in finding insights for our clients.
Custom GPTs
What I don't use much are custom GPTs because I didn't find a use case for them. After AI Restart I have some new use cases where this tool should be helpful in my daily work.
The first use case is to use them as your internal tool for your coworkers who should be able to train with ChatGPT. You should predefined how to respond - For example, you should make a tool that helps them to analyze their prompts and return what they should improve.
The second use-case is that you should upload into Custom ChatGPT your brand manual and other pieces of information about your brand and after that, you should make copy/content that will be in your tone of voice because GPT should extend their knowledge (your prompt) with this manual. So this should be better than prompting in classic ChatGPT and trying to pass on the experience you need for the output you are lookfor.
Usage o in CX
Also, there was a great showcase of how should be used LLM in the process. One of them that I really liked was about customer experience in which the speaker has shown that you should delegate a lot of tasks on LLM in your customer service team.
From responses to typical questions about where is my package to resetting passwords. These tasks should be done by LLM. Also, LLM should respond to questions about your products and offer interesting deals or up-sell or cross-sell. And also you should use AI in your team to get them the right pieces of information faster.
Speakers also showed us some synthetic voices which yes should be detected for now. But in two or three years. It won't be possible to identify if you are talking with a person or a bot and a lot of people should do much more interesting jobs.
Let LLM think about and write OK
Another presentation that I want to mention was “denying” some information about LLM that should be clear to all and I think that it was but there are still a lot of people that don't understand how LLM works. So for me, this was a great showcase of how to “talk” with friends or family who aren't enough involved in LLMs.
For example, there was a great “show” in which three people on stage “show” how LLM responded. They have only one word to make a poem and it doesn't work well because they won't be synced. So the main information is that what LLM didn't mention in the text that didn't know.
So If you write to LLM “Let's think about … and when you are ready then respond OK” then it responds only OK without any thinking in the background because the goal is to write ok and ok will be probably one of the highest probable worlds after this sentence. 🙂
And that's all here for everyone. If you are a reader from Czech Republic then yes - it should be interesting to attend AI Restart. Some presentations won't be “enough” for me but the rest of the conferences bring great tips that I should use. And I will be happy if you join me here on Friends of Advanced SEO to get more tips from AI restart.
What I prepared for paid readers?
Automatization marketing with GenAI and connection to GCP
Automatization with AI on project management
Company AI transformation on steroids
Aren't we controlling LLM too much?
If you are from the Czech Republic then I have another small promo here. In two or three weeks I have the first workshop in which I will talk about practical usage of LLM in my daily flows.
So you should look at it on my personal website. Maybe it will be translated with AI into English. 🙂