Operationalizing generative AI for SEOs. A look at how generative AI models work. The five lelves of prompt engineer. How to build and deploy a practical GenAI application and what is the future of AI for marketers. The presentation is given by Dave Davies who started SEO in 1999. In 2004 he founded a SEO company and he started covering machine language in SEO articles back in 2016. It's 64% certain that he is a person.
1 of 25
Download to read offline
More Related Content
A presentation on Operationalizing generative AI for SEOs
2. Got my start in SEO in 1999.
Founded an SEO company in 2004.
Started covering ML in SEO articles in
2016.
Started covering ML at conferences in
2018.
64% certainty that Im a person.
Who am I?
3. A look at how generative AI models
work.
What am I covering?
The five levels of prompt engineering.
Building and deploying a practical
GenAI application.
The future of AI for marketers.
4. What am I covering? The prompt:
Create a cartoon version of a bee with a magic wand . The bee is crossing items off a
chalkboard . The chalkboard items include the following list, which are the topics to be
included in a presentation the bee is giving on AI.
A look at how generative AI models work.
The five levels of prompt engineering .
A look at some practical examples of generative AI applications.
The future of AI.
The nightmare fuel
that Google
ImageFX (Imagen 2)
came up with.
6. How generative AI models work
Training:
Garbage in. Garbage out.
Tokens:
Words? We dont need words.
7. How generative AI models work
Training:
Garbage in. Garbage out.
Neural Networks:
Like your brain, but not.
Tokens:
Words? We dont need words.
8. How generative AI models work
Training:
Garbage in. Garbage out.
Neural Networks:
Like your brain, but not.
Output:
Right or wrong, its about probabilities.
Tokens:
Words? We dont need words.
Catch Ryans session
Will AI Overviews kill SEO
tomorrow at 8:50am.
9. The five levels of prompt engineering
From theTELeR paper: https://arxiv.org/pdf/2305.11430
10. Five levels of prompt engineering: level one
A simple one-sentence directive expressing the high-level goal.
Example
Write a Tweet about AGI.
Output
11. Five levels of prompt engineering: level two
Multi-sentence directive expressing the high-level goal and the sub-tasks that need to be
performed to achieve the goal.
Example
Write a Tweet about AGI. It should be:
- Supportive of AGI
- Acknowledge the concerns of others around safety, privacy and security
- Focus on the benefits to humanity
- Use examples from health and agriculture
Output
12. Five levels of prompt engineering: level three
Complex (bullet-style) directive expressing the high-level goal along with a detailed bulleted list of
sub-tasks to be performed.
Example
Write a Tweet about AGI. It should answer the following questions:
- What is the biggest benefit for humanity?
- Why is safety not a concern?
- How can we avoid job displacement?
- How can we offset the environmental impact?
The Tweet should be:
- Supportive of AGI
- Acknowledge the concerns of others around safety, privacy and security
- Focus on the benefits to humanity
- Use examples from health and agriculture
It can be a Tweet thread.
Output
13. Five levels of prompt engineering: level four
A complex directive that includes the following: 1) Description of the high-level goal, 2) A detailed
bulleted list of sub-tasks, 3) A directive for the LLM to explain the output.
Output
Example
Write a Tweet about AGI. It should answer the following questions:
- What is the biggest benefit for humanity?
- Why is safety not a concern?
- How can we avoid job displacement?
- How can we offset the environmental impact?
The Tweet should be:
- Supportive of AGI
- Acknowledge the concerns of others around safety, privacy and security
- Focus on the benefits to humanity
- Use examples from health and agriculture
It can be a Tweet thread.
Explain clearly how the thread provides a compelling case for pursuing AGI
research.
14. Five levels of prompt engineering: level five
A complex directive that includes the following: Everything in level four, plus the addition of
guidelines on how the output will be evaluated and/or few-shot examples.
Output
Example
Previous message plus:
The language, style, tone and structure should be inspired by these three top
performing Tweets by the same company.
Join @maxjaderberg & @SergeiIakhnin from @IsomorphicLabs to explore
how machine learning accelerates biotech and drug discovery.
稲 ю: https://lnk.to/GDpc9T8b #GradientDissent
Introducing Weave: a suite of tools to develop and productionize AI applications.
Learn more here URL Join @altryne, to understand the problems Weave is
helping you solve!
Our latest blog reviews the GPT2-chatbot. Learn about our tests, insights on its
potential, and how it is comparable to GPT-4. Read more URL
(@BrettYoung3d)
20. The future of AI on the marketing sector
So, you cant keep doing what youre
doing or you will be replaced by AI.
21. The future of AI on the marketing sector
Whats coming soon will undoubtedly
look something like this.
22. Were using Generative AI to
Write social posts
Outline or even write content
Generate schema and analyze
technical issues
Create paid copy
Analyze data
Creating and personalizing email
Etc
The future of AI on the marketing sector
23. Questions to ask yourself
If everyone uses the same model,
who wins and how can it be you?
What training do you need to be one
skill ahead?
Who is your competition?
(hint: its not AI)
How can you be the bridge between
different AI systems and functions?
The future of AI on the marketing sector
24. Resources
X: @onlineinference
TELeR paper: https://arxiv.org/pdf/2305.11430
GEO article: http://wandb.me/GEO
Blog: https://wandb.ai/onlineinference/
Company: https://wandb.ai/
Thank you!
Now there only a 50%
Chance that Im a person. :D