Football Marketing & AI’s Achilles Heel
How about those Bears!
Maybe they should have thrown in some AI Robots like the ones that showed up to the Dolphins / Chargers game.
A better football marketing story coming out of Batavia high school.
We're also covering how Large Language models work and where their blind spot is and how to exploit that.
Part 2 on building your own AI Chatbot
And, if we have time, using ChatGPT to code something I never could have done on my own for a site staging an arts for mental health day across 18 cities.
~subheadIndex-links
Changes In Format
Recording are now accessible without login
This is teaching me a lot by focusing on teaching you
Adding podcast this week
No longer doing a Google Doc agenda — everything is going to these pages
Membership will be required for a free copy of my marketing platform
Weekly day and time changing to accommodate interns
Apologies for missing last Friday Office Hours
Watch the full video session
Listen to the full audio of the session
Outline of the session
I. Intro - Football
AI robots at NFL game as marketing stunt for movie "The Creator"
Shares marketing story from Batavia High School football team
Freshman student Sierra made great TikTok videos interviewing players
Allows audience to get to know players in authentic way
Demonstrates power of short videos for engagement and storytelling
II. Changes in Bootcamp Format
Moved bootcamp library from membership system to public site for accessibility
Putting full agenda and notes on bootcamp pages rather than Google Docs
Recorded sessions posted on pages with podcast version coming
III. Mission of Bootcamp
Created to help clients, interns, teach AI and marketing tools
Connect local companies with interns
Shift to short "TikTok-style" videos and AI tools
IV. AI Robots at Football Game
Discussion of AI robot marketing stunt at NFL game
Asks if robots look real or not
V. Football Marketing at Batavia High School
Details marketing initiative led by student intern Sierra
She made engaging TikTok videos interviewing players
Allows audience to connect with players
Power of short, authentic videos for engagement
VI. New Football Website
Showcases new Squarespace site made for school's football marketing
Template designed for video and images to showcase Sierra's videos
Goal to expand site for other sports/clubs at school
VII. Building Your Own Chatbot
Demo of using Zapier to build AI chatbot
Connect chatbot to data source and chat with customized responses
Still early stages but important capability for custom chatbots
VIII. Review of Bootcamp Structure
Recaps 50 min session with 10 mins for questions
Reminder of Friday office hours
IX. Facebook Groups for AI Tips
Recommends Facebook group "MidJourney Prompt Tips" to learn about prompts
X. Past Office Hours: MidJourney Practice
Recaps collaborative MidJourney session at office hours
XI. Video Editing Tools
Recommends Kapwing for easy browser-based editing
Also suggests using Descript for transcription, cleaning up audio
Two tools can be hard for beginners - start with just Kapwing
AI Robots
Miami Dolphins game against the Los Angeles Chargers — AI robots for an upcoming movie called “The Creator:”
"The Creator" is a science fiction action thriller film directed by Gareth Edwards. The movie is set in a future where humans and AI are at war. The plot follows Joshua, an ex-special forces agent who is recruited to hunt down and kill the Creator, the elusive architect of advanced AI who has developed a mysterious weapon with the power to end the war and humankind itself
Everyone is debating whether they are real or not
Batavia Football — A Better Marketing Story
Freshman jumping in with their phones, external audio and asking questions
Coach Piron — everyone knows him, but most outside the students don’t really know him — now they do
The Squarespace Template
https://bhsfootball.squarespace.com/
Password: bhs
News: Meta Is Developing a New, More Powerful AI System as Technology Race Escalates
Meta (parent company of Facebook and Instagram) is developing a new AI system intended to rival OpenAI's GPT-4 model. The goal is for it to be ready by early 2024.
This new model would be several times more powerful than Meta's current Llama 2 model released in July 2022.
The system is being developed by a team formed by Mark Zuckerberg earlier this year to accelerate Meta's generative AI capabilities.
Meta plans to train the model on its own infrastructure, unlike Llama 2 which was trained with Microsoft. Meta aims to open source the model when complete.
The move is part of Zuckerberg's push to position Meta as a leader in AI amidst intensifying competition with companies like OpenAI and Google.
There are potential benefits but also risks to developing such a powerful open source AI model, including issues around copyright and content moderation.
Meta's model aims to match GPT-4's capabilities, though it may not fully close the gap with rivals given their progress. The race to develop more advanced AI continues.
News: OpenAI Launches Business Version of ChatGPT That Competes With Microsoft
From WSJ: 28 Aug 2023
OpenAI has launched ChatGPT Enterprise, a business version of its popular ChatGPT chatbot. This puts OpenAI in direct competition with Microsoft, its biggest backer and partner.
ChatGPT Enterprise is tailored for corporate users - it can help employees learn skills like coding, analyze internal data, and runs twice as fast as the regular ChatGPT.
Microsoft recently launched a similar product called Bing Chat Enterprise. Both tools promise to keep customer data private.
OpenAI and Microsoft have had a close partnership, with Microsoft investing billions in OpenAI. But there has also been some conflict between the companies.
OpenAI says ChatGPT Enterprise moves towards an "AI assistant for work" while protecting company data. Microsoft describes Bing Chat in similar terms.
This is OpenAI's first ChatGPT product tailored for business. It already sells access to its models to companies.
The launch comes amid an ongoing shortage of advanced chips needed to run AI systems. OpenAI focused on maximizing speed and reducing latency for ChatGPT Enterprise.
How Large Language Models (LLMs) Work
Here is a detailed outline of the key points in the transcript:
I. Introduction
- This is the final episode in a 3-part series on A.I.
- The focus is on what the world will look like when A.I. is everywhere.
II. The Job of "Prompt Engineer"
Anna Bernstein was an early pioneer in the new role of "prompt engineer"
Prompt engineers help users communicate effectively with AI systems by:
Crafting precise prompts that translate goals into instructions AI can follow
Understanding nuances of human communication:
Humans rely heavily on shared context and "fuzziness"
AI models interpret prompts literally
Finding the right words and syntax to "unlock capabilities" in the AI
Editing prompts obsessively until AI output matches intent
Bernstein's background in poetry helps with this linguistic precision
She has spent countless hours "talking" to AI models to understand their capabilities
Prompt engineering requires:
Identifying the exact purpose and audience for output
Using powerful action verbs to convey goals
Avoiding negatives, which can confuse models
Adding lots of synonyms to emphasize style/tone
This helps models produce more dynamic, nuanced results
Prompt engineering has become an in-demand job, indicating AI's business relevance
Bernstein's skills may become applicable across industries as AI use grows
III. How A.I. Models Learn
Dario Amodei provides an overview of how large language models (LLMs) like GPT learn:
LLMs start with no knowledge, like a newborn baby.
They are fed massive amounts of text data from the internet - trillions of words.
The LLM tries to predict the next word in each document, learning to match patterns.
At first predictions are random, like a baby babbling.
Over time the LLM learns:
Basic spelling and grammar
How to structure sentences grammatically
Common facts about the world (e.g. Paris is the capital of France)
More advanced concepts like math and computer programming
Humans can't manually track everything a LLM learns. But they can test it by asking questions.
After basic training, humans intervene to:
Remove toxic content and align with human values
Craft a "constitution" that codifies principles for the LLM to follow
Have the LLM monitor itself for constitutional alignment
This human involvement is essential to prevent harmful outputs from raw LLMs.
Ongoing research focuses on developing LLMs that learn "the right stuff" from the start.
IV. Benefits and Dangers of A.I.
- Amodei sees huge potential upsides of A.I. in fields like medicine.
- But also serious risks like bioweapons if not properly regulated.
V. Using A.I. as a Thought Partner
- Ethan Mollick finds A.I. helpful for entrepreneurship by overcoming inertia.
- He demonstrates "constrained ideation" to generate business ideas with A.I.
- The process helps make your own thinking clearer.
VI. A.I. and Education
- A.I. could help education by creating tests and customizing learning.
VII. Conclusion
- Key takeaways:
- Current A.I. is not worthy of the hype.
- Its future impact depends on how humans shape it.
- It will transform jobs - prepare by understanding it.
- Anecdote: A.I. helped a man say goodbye to his dying wife.
What is AI chatbot's achilles heal?
AI chatbots need large amounts of data to learn from and to generate responses. However, not all data is accurate, relevant, or unbiased. Some data may contain errors, inconsistencies, or misinformation that could affect the chatbot’s understanding and output. Some data may also reflect the biases or prejudices of the data sources or creators, which could lead to unfair or unethical outcomes for some users or groups.
Twitter has blocked AI Scraping
Twitter has blocked chatbots from scraping data. Twitter has taken several measures to prevent AI-based chatbots from scraping data, including blocking non-users from viewing tweets unless they are logged into an account , restricting the platform's data access to ChatGPT web browser extension , and blocking data access to AI-based chatbots like ChatGPT 3 . Additionally, Elon Musk-owned X, formerly Twitter, has updated its terms of service to prohibit scraping and crawling. Elon Musk has also stated that Twitter users will have to temporarily sign in to view tweets to stop AI data scraping, and Twitter will place limits on the number of tweets people can read per day 5 . Twitter CEO Elon Musk has also mentioned that unregistered data scrapers have been slowing the service down.
IPS Teacher Interviews:
There were things said that no words written yet can convey
Topic: Your Own AI Chatbot
Using Zapier’s new Interfaces tool, I’ll show you have to build your own Chatbot that uses your own data.
Connect a custom data source
Upload one 1MB static text file (.txt, .csv, .json, .pdf, .doc, .docx).
a plain ASCII English text file without compression could hold approximately 150,000-200,000 words per megabyte.
300-400 printed pages in a Google doc.
