The splashy release of ChatGPT from OpenAI has kicked of 2023 as the year of Chat Bots. Other major players are hustling to bring out their own AI backed chat bots, whether anyone wants them or not.
This month, Microsoft and Google presented their own bots [2]. Microsoft “Bing Chat” is in Beta and, as the name suggests, connected to their Bing search engine. Google is working on something they call “Bard”, that will presumable work with Google Search,. Bard is coming “Real Soon Now”.
ChatGPT has provided the internet and chattering classes with abundant fodder for both ridicule and serious concerns. ChatGPT strikes me as “not ready yet”. But the competition, makes ChatGPT look positively sound.
“both Bard and Bing Chat stumble out of the gate”
(from [2])
Bing Chat could be tried, but the results have been mystifying. Trial users comment that Bing Chat seems “combative and defensive”, and not only makes up false facts, it makes up fake references, which it cites to back up false facts. Zounds!
Google’s Bard isn’t ready for users, and the video demo features a blatantly inaccurate answer. Everyone wonders why Bard didn’t do a Google search, which would have returned the correct answer!
These bots all use similar technology, though they have different training sets and have managed to create somewhat distinctive “personalities”. While these artificial personalities may be entertaining to play with, they don’t seem to be particularly promising as sources of useful information.
I mean, the Internet is already a vast sea of untrustworthy and made up junk. Do I need my search assistant making up its own facts on top of everything else? [1]
Perhaps the most interesting point Smith makes is financial. We still haven’t got much idea what the business model might be for these liar-bots. I gather that ChatGPT is trying $20 per month ($240 / year!) subscription for the current version. The others aren’t even ready to talk about money.
If these fabricator-bots are supposed to be a “simple, friendly” interface to searching the web, what would the revenue model be? How much will people pay for wrong answers? Search is currently funded through advertising, so maybe these BS-engines might embed ads into the BS they generate? That sounds horrid to me, but we’ll see.
The question of potential revenue is no small thing, because the AI behind these bots is pretty darned expensive to create and run. Smith reports estimates that scaling up ChatGPT to what Google search does would cost tens of billions of dollars [1]. If this is anywhere near correct, just how could such expenditures be recooped?
“deploying OpenAI’s ChatGPT at the scale required by Google Search would lead to US $36 billion dollars in new recurring costs and could require over $100 billion of new server and network infrastructure.”
([2])
Phew! That’s a lot to pay for something that screws up your advertising revenue, doesn’t even work, and no one actually asked for.
- Stephen Shankland, Computing Pioneer Criticizes ChatGPT AI Tech for Making Things Up, in cNet, February 18, 2023. https://www.cnet.com/tech/computing/computing-pioneer-criticizes-chatgpt-ai-tech-for-making-things-up/
- Matthew S. Smith, Google and Microsoft Race to Unveil AI-Based Search, in IEEE Spectrum – Artificial Intelligence, February 20, 2023. https://spectrum.ieee.org/chatgpt
Robert McGrath's Blog 
One thought on “The Chatbot Race”