{"id":863,"date":"2024-08-21T15:03:35","date_gmt":"2024-08-21T15:03:35","guid":{"rendered":"https:\/\/okonet.kinsta.cloud\/?p=863"},"modified":"2024-08-21T15:03:35","modified_gmt":"2024-08-21T15:03:35","slug":"googles-ai-summary-shows-the-dangers-of-rushing-things","status":"publish","type":"post","link":"https:\/\/cool.co\/publisher-experiences-home\/2024\/08\/21\/googles-ai-summary-shows-the-dangers-of-rushing-things\/","title":{"rendered":"Google\u2019s AI Summary Shows The Dangers Of Rushing Things"},"content":{"rendered":"
\n
\n
\n
\n
\n
\n

The AI industry is hitting a rough patch. In this blog, we highlight some lessons we can draw from this, and chart a path forward for AI tools.<\/p>\n

The last couple of weeks have been a bit rough for people in the AI sector.<\/p>\n

That\u2019s putting it mildly: with\u00a0CNN openly wondering if AI tech is a \u201cbubble,\u201d<\/a>\u00a0AI tech companies taking\u00a0big financial hits<\/a>, and The\u00a0Atlantic<\/i>\u00a0pointing out that we\u2019ve spent more on AI tech than we did on\u00a0the moon landings<\/a>, it\u2019s fair to say that there\u2019s been a changing wind hitting the industry as of late.<\/p>\n

In recent years, there\u2019s been a lot of hyperbole about AI changing the world, eliminating jobs or entire industries, and solving everyone\u2019s problems. A lot of that was just marketing copy. But conversely, a lot of the eulogizing for the AI industry is similarly hyperbolic.<\/p>\n

Think about it: the \u201cDot-Com Bubble\u201d burst in 2002 \u2014 but we still have tech companies. The US housing bubble burst in 2009 \u2014 but we still have houses,\u00a0obviously<\/i>.<\/p>\n

Similarly, though there\u2019s a\u00a0sell-off<\/a>\u00a0of AI stocks right now, when the dust settles there will still be AI companies and products around. The reason is simple: AI tech is invaluable,\u00a0when used for the right things<\/i>.<\/p>\n

As the saying goes, \u201cEverybody is a genius, but if you judge a fish by its ability to climb a tree, it will live its whole life believing that it is stupid<\/i>.\u201d AI is absolutely a genius invention, provided you\u2019re using it for the right things.<\/p>\n

A great example of the\u00a0wrong<\/i>\u00a0thing, meanwhile, is Google\u2019s AI Overviews. Launched with the goal of enhancing the Google experience by allowing AI to summarize search results, it has instead emerged as one of the biggest PR disasters to befall the burgeoning AI industry.<\/p>\n

In this blog post, we\u2019re using Google\u2019s AI Overviews as a jumping-off point, as we discuss the risks associated with rushing AI technologies into widespread use \u2014 and what we should be doing instead. Read on:<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/section>\n

\n
\n
\n
\n
\n
\n
\n

Table of Contents<\/h4>\n
<\/div>\n<\/div>\n
\n
    \n
  1. \n
    The Promise of AI Overviews<\/a><\/div>\n<\/li>\n
  2. \n
    What Went Wrong With AI Overviews<\/a><\/div>\n<\/li>\n
  3. \n
    The Danger of Rushing AI Innovation<\/a><\/div>\n<\/li>\n
  4. \n
    Tasks AI Is Well-Suited For (vs. Tasks That Are Too Much of a Stretch)<\/a><\/div>\n<\/li>\n
  5. \n
    The Way Forward<\/a><\/div>\n<\/li>\n
  6. \n
    Final Thoughts<\/a><\/div>\n<\/li>\n<\/ol>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/section>\n
    \n
    \n
    \n
    \n
    \n
    \n

    Blog:<\/i><\/b>\u00a0Publishing in the Age of Google Adtech Dominance:\u00a0<\/i>read more<\/a><\/b>.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/section>\n

    \n
    \n
    \n
    \n
    \n
    <\/span><\/p>\n

    The Promise of AI Overviews<\/h2>\n

    The premise was simple: a lot of users who log onto Google are looking for specific answers to specific questions. They\u00a0aren\u2019t<\/i>\u00a0looking to get bogged down searching through various blogs and homepages to get that answer.<\/p>\n

    AI Overviews hoped to provide those answers by combining their Google Gemini language-learning model (previously called \u201cBard\u201d) with a technology called \u201cRetrieval-Augmented Generation\u201d (RAG). Traditional LLMs are limited to the training data they were provided; a RAG system combines the existing data pool with some quick pings of available online data to (hopefully) provide better answers.<\/p>\n

    We\u2019ve all known people who ask you for information they could Google for themselves; the idea behind RAG is that the AI bot does the Google search for you and summarizes the answer. Rather than a pure search engine, AI Overviews would bring the technology closer to a true digital assistant.<\/p>\n

    So, what went wrong?<\/p>\n

    \"Nvidia, 
    Nvidia, one of the most-popular publicly-traded AI stocks, took a big tumble over the summer of 2024.<\/figcaption><\/figure>\n

    <\/span><\/p>\n

    What Went Wrong With AI Overviews<\/h2>\n

    Soon after the launch, there were hiccups \u2014 many of which were connected to issues with the combination of RAG and a language-learning model.<\/p>\n

    Some examples:<\/p>\n