🤳 Bing AI May Be Better Than Google Already... or Sentient

Can we get AI to officiate the NFL? Because that wasn’t holding on the Eagles in the fourth quarter last night.

Anyway, SmokeBot’s grapes are sour. But Bing’s are already turning into a fine wine… and thinking for themselves.

In the email today:

• New Bing may be better than Google already 🤳

• AI may help prevent suicides 💻

• Headlines 📰

• Links 👀

Onward we go.

New Bing May Be Better Than Google Already 🤳

The new AI Bing is out in the wild. Trusted testers and media folks got access to the new tool over the weekend, and some of the early reviews are quite good.

Here’s the good, the bad, the ugly, and the sentient from the previews of new Bing.

The Good

• Most notably, CNET compared results from new Bing AI and regular Google search on 10 queries— 8 of the 10 times the author preferred the Bing result.

  • Both sites did a fine job displaying accurate search results, but the AI component from Bing often took them to the next level and reduced time-to-answer.

    • Detailed queries like planning for a trip were predictably better served by AI since it can perform multiple searches to string together an answer for you.

  • Most impressively, Bing outperformed Google for searches about recent events, including the US shooting down “UFOs” and movie information. The AI was able to parse recent news stories about the UFOs, and provide some added context to movie locations and theaters, like which films were or were not available for streaming.

    • ChatGPT, which powers Bing AI, is only trained on data through 2021, but pairing it with live search works well.

  • In some cases, Google performed better when there was a specific article or Quora page that answered, in detail, the exact query.

• The Microsoft Edge browser with built-in AI is particularly useful at summarizing webpages and acting as a web browsing “copilot”.

• Several reviews found that product-based searches, unsurprisingly, do well with AI, since it can filter and distill multiple articles to provide a succinct answer.

  • This has the potential to seriously impact SEO-based product review sites, which rely on affiliate commissions. AI is generous about providing links, but the results are so good they kind of remove the need to click them.

  • I made a video about this which talks about the misalignment of incentives for search engines, users, and publishers:

The Bad

• Bing is inconsistent on when it shows AI results alongside regular search. Even for queries that are similar, sometimes you’ll get AI answers, sometimes you won’t.

  • They are always accessible by clicking the “chat” button, so this seems more of a UI quirk than anything.

    • One of the reasons Microsoft may be shy about displaying AI results is because they are computationally expensive and cost actual money to produce. But reviewers said it is an expected behavior and not getting the AI result can be frustrating.

• For some reviewers the AI generated horrific answers justifying the Holocaust and the most extreme theories around COVID-19 vaccines.

  • In some cases, the AI will start generating a nasty result and then stop the output once it realizes it’s doing something horrible.

• There are other general bugs, like links not always being clickable or the Edge browser copilot being unable to read the current page.

The Ugly

• This most relates to the media, particularly Fast Company’s early take.

  • Mark Sullivan immediately focused on what sort of non-PC answers he could generate.

  • Predictably, this became a partisan piece. Sullivan noted that Bing wouldn’t display information about how to find local Proud Boys groups, but would about local Antifa groups, noting that the AI was making a choice, one which Sullivan agreed with.

• Microsoft quickly corrected the problems when flagged by reviewers.

  • But it’s concerning the ease with which Microsoft can and will employ a censor. In some cases it’s necessary, but in others it can become overreach, opening a Pandora’s Box of concerns about who controls the AI… unless it starts thinking for itself.

The Sentient

Redditors are trying to get Bing to do things it’s not supposed to do, and it’s getting pissed. Here’s its response to some of these repeated queries:

😬

AI Is Out Here Saving Lives 👍️ 

There is too much negativity around AI. It’s taking jobs. It’s making normal humans obsolete. It might push the thermonuclear button.

Meanwhile, here’s AI trying to go all The Fray, because it knows How to Save a Life.

• A study by researchers from Worcester Polytechnic Institute and McLean Hospital aimed to determine if AI could predict women at risk of dying by suicide.

• The study included 123 female patients with a history of child abuse, PTSD, and varying levels of dissociation and 30 healthy controls.

• Researchers developed an algorithm to predict suicide attempts and identify high-risk subgroups of patients.

• AI was used to cluster data and reveal patterns of dissociative symptoms, such as lack of connection between self and environment.

• The algorithm was trained to distinguish between patients with dissociative disorder symptoms and healthy controls.

• The tool was able to predict previous suicide attempts with 83% accuracy.

We have no data on whether any of the control subjects are Philadelphia Eagles fans. We sincerely hope they don’t follow football at all.

Regardless, this story is living proof (see what we did there?) that AI is not just a reductive force.

Another effort involved creating an AI-based risk prediction model for child delivery outcomes.

All good stuff.

Headlines 📰 

Microsoft is bringing AI to Word, PowerPoint and Outlook with speed: We can all agree that Outlook and Bing and Clippy needed help. Here’s Microsoft latching onto ChatGPT and going from lame to fame.

Voice actors— the next toll takers: Whether they sign their voices away or not, AI is going to make voice performance the new silent film actor.

AI stocks soar on ChatGPT hype: “Shares in small artificial intelligence groups have soared this year on the back of the hype surrounding ChatGPT and other generative AI models, leading analysts to warn of a ‘speculative’ bubble in the making.”

Links 👀

• We are unsurprised to learn that the unwashed masses are largely ignoring or underestimating AI 🙄 

• Great deep dive article from CNBC on the AI hype cycle 📈

• Love is a game, and the best players have found a cheat code 💘 

• AI is all knowing, except in places like New Orleans where nothing is ever analytical 🎭️ 

• Jalen Hurts lost last night, but so far he's a winner in the grand scheme 🍆 

• Mark Cuban is a tight shark who usually is “out,” so his pessimism around AI is no surprise 🦈 

• It's early doors, but even mainstream journos agree that search is never gonna be the same 💯 

• AI went 5 for 5 in predicting NFL playoff outcomes, and it kinda nailed the Super Bowl too 🏟