Lesson#4: Wikipedia vs AI (Part 4)

This is the fourth part of a summary of an article in the New York Sunday Times Magazine (7/23/23) about how Wikipedia is being used by AI chatbots for their current results and success.

The last time I saw anything catch on as the current AI Chatbot phenomenon was when I first watched the beginning of the internet, then e-commerce, then social media, and then online teaching. There seems to be no point to fight its use because all of the big tech companies are daily piling money into it.

CURRENT    ROSTER    OF    CHATBOTS

• The leader: ChatGPT

• The runner-up: Google Bard

• Open source: HuggingChat

• For building your own chatbot: Zapier AI Chatbot

• For searching the web: Microsoft Bing AI, Perplexity, YouChat, KoalaChat

• For content writing: Jasper Chat, Chat by Copy.ai, ChatSonic, ZenoChat

• For sales and marketing: ChatSpot

• For messaging: Personal AI

• For personal coaching: Pi

• For tinkering: OpenAI playground, Poe, DeepAI Chat

• For fun: Character.AI

• On social media: Snapchat My AI

• For coding auto-complete: GitHub Copilot, Amazon CodeWhisperer

• Up-and-coming: Claude, Khanming

And when I say piles of money, I am saying billions of dollars even though there are many flaws in the development process. Meanwhile, Congress is trying the catch-up game desperately trying to understand it and considering several bills to regulate the marketplace.

In April '23 Reddit announced that it would not make its content available for "scraping" by big tech companies without being compensated. It's doubtful that Wikipedia can do the same because of its terms of service, but Gertner suggested that it could ask all of the chatbots to give them higher attributions and offer citations in their findings. And this is because everyone Gertner spoke to in the tech community while researching this article said they don't even think AI is possible without Wikipedia.

This is because AI researchers have looked at the scenario of whether it is possible for AI Chatbots to be able to train fledging, new chatbots. The researchers found that using current robust AI chatbots to train "new" chatbots led to chaos. Ilia Shumailov, one of the researchers from Oxford University, said, "Without human data to train on, your language model starts being completely oblivious to what you ask it to solve, and its starts talking in circles about whatever it wants, as if went into this madman mode." Basically, the study concluded that the value of human generated data (Wikipedia) will be increasingly valuable for future AI large language models.

Gertner concluded by saying that AI systems need to understand the theoretical process of human alignment. Alignment basically is an understand of the human nature of content and it is an enormous priority for AI systems because a system out of sync with human nature might create terrible damage.

Getner end the article by quoting Jade, a Wikipedia editor who has 24,000 edits to her credit. "we are going to have to create processes, we are going to have to have hard conversations about the ethics of AI to create Wikipedia articles. And I don't ever --- maybe not never, but certainly not in this century do I see robots fully replacing humans on Wikipedia."

                                                   --- End of Wikipedia vs. AI Part 4 of 4 ---

                                                                                # # # #

Source: New York Times Magazine Article: Moment of Truth - Can  Wikipedia help teach AI chatbots to get their facts right, Jon Gertner, 7.23.23, page 34.