Midway
Como
Frogtown
Como
Frogtown
News for everyone, delivered to everyone
“There’s going to be a huge impact on media and the culture based on Generative AI,” said Danny Olson during a Midway Area Chamber of Commerce luncheon on Oct. 16, 2024.
He’s the AI and Emerging Platforms Lead at Weber Shandwick, a global communications agency, and his background includes sales, digital marketing, and public relations.
Olson pointed to the new Google overview summaries generated by AI that are appearing at the top of search pages that can pull information from a variety of sources. They can be read without clicking into another source, which means people won’t be going to other businesses or media websites, using their information without paying for it. AI is bringing new gatekeepers, he pointed out. “It dictates how we learn about the world and go about our day.”
Olson quoted former Google CEO Eric Schmidt: “We’re going to have a very different world and it’s going to happen very quickly. Imagine if you had Aristotle to consult with you on logic. Oppenheimer to consult with you on physics. Not the person but rather the knowledge, the scaling intelligence of these truly brilliant people who were historically incredibly rare. What are we going to do when super intelligence is broadly available to everyone?”
AI CHANGES EVERYTHING
“Let’s talk about how this changes everything,” remarked Olson.
This changes search.
“We’re not even getting to the click,” he said, pointing to how this will have a huge impact on media.
It changes learning.
Olson observed that right now he’s using generative AI to write about 80% of his performance reviews. He puts all the feedback and information into AI, which then generates a summary, saving him hours of time.
It changes media.
People can create reporters, background and voices that are entirely generated by AI.
It changes influencers.
There are now individuals on social media who are not real. Their photos and the accompanying text are completely created by AI.
It changes experiences.
Researchers at Shenzhen People’s Hospital in China using an AI fusion of multisource data to identify key features of vitiligo, a skin condition characterized by the loss of skin pigment, and to help with its diagnosis and treatment.
It changes risk.
“It changes us,” said Olson.
He acknowledged that there is a great deal of fear around AI right now and people are worried that AI will replace people in jobs.
His personal belief is that there won’t be fewer jobs but that there will be different jobs for people. “You and technology working together,” Olson said, “not you versus technology.”
What is generative AI?
Generative AI maps patterns and relationships in large data sets, and uses that knowledge to create new content, such as images, video or text, explained Olson.
What is AI good at now?
• Producing coherent and contextually relevant text.
• Translating text with high accuracy.
• Creating images and art from descriptions.
• Analyzing large datasets for patterns or trends.
• Summarizing documents into different formats.
• Tutoring and explaining a wide range of topics.
• Generating new thoughts for ideating and brainstorming.
• Generating creative content like art, music and fiction.
AI is useful not for answers, but for ideas, said Olson. “It is meant to be iterative where you go back and forth with it, versus expecting a single polished answer.”
Where does AI struggle?
• Suffers from vulnerabilities in data privacy.
• Encounters difficulties performing basic math.
• Limited in creating truly original ideas or concepts.
• Biases present in its training data.
• Liabilities due to complex AI ownership laws.
• Struggles with recent events post-training cut-off.
• Hallucinating nonsensical or irrelevant results.
• Inaccurately generating images of specified text.
• Lacks deep understanding in specialized fields.
• Struggles with adhering to specific word counts.
“AI will give you an answer but it doesn’t always have the right information to give the answer,” observed Olson. “That’s where the interrogation comes into play.” AI is a collection of data but it doesn’t necessarily have all of the data on a subject.
“It’s not truly creative,” he added. “It can only imagine what exists.”
WHERE TO START?
Current AI platforms include:
• SENSING: Blackbird.ai and NextAtlas
• SEARCH: Perplexity, ChatPDF, Opinionate.io, and Consensus
• CREATION: Adobe Firefly, Leonardo.ai, DALL-E, Runway and ChatGPT
Olson pointed out that there is a typical learning curve that moves beyond the hype. First, people are excited that ChatGPT can answer all their questions. Then they realize that ChatGPT doesn’t understand what it is saying and is only predicting plausible answers. They realize that ChatGPT will return incorrect answers that are statistically plausible but not based in fact. They begin to understand that ChatGPT is best used when there isn’t a precise “right answer.” Only then do they recognize how to get value from ChatGPT and avoid the risks.
A mindset shift is required. AI isn’t a silver bullet.
This makes the “prompt” really important.
“We are serving as editors or as people engaging with the product,” said Olson.
The process should involve a prompt from a human that is detailed with clear instructions for a specific task.
It isn’t about giving one prompt and getting one answer. It’s about doing a prompt, getting an answer, doing another prompt, getting another answer, and doing that multiple times before a human edits and creates the final human output.
Olson’s mom is an esthetician and small business owner. She isn’t a marketer. But she can use AI to generate email subject lines.
There are also legal issues about taking information directly from a prompt and using it as is. All of the information in AI was pulled from someplace else. The legal consensus, according to Olson, is that you can’t take stuff from a prompt and use it directly “much like you wouldn’t take a piece of text from Google search and reprint it as your own.”
It’s tricky right now because generative AI isn’t providing a source for where the information came from.
“Assume you are using copyright material,” Olson recommended.
The other issue with AI material is that the programs are biased. “They are built by people,” said Olson. “There might not be explicit bias but [there is] implicit bias. Don’t take it word for word.
“The interactive use of it is how you get the most value.”
Comments
No comments on this item Please log in to comment by clicking here