In yet another round of corporate ours-is-bigger stat-waving, Facebook’s new Blender chatbot has thrown down the gauntlet to Google’s Meena. But in reality, most businesses won’t ever need these super-AI bots for their customer-facing or internal processes.
The technology industry is prone to bouts of statistical willy-waving. From AMD vs Intel when it comes to processor speed or number of cores, Microsoft vs Apple on OS features, any smartphone vendor over megapixels in their cameras and so on. Now, the cloud-era heavyweights are at it, with Facebook taking a pop at Google over their future-ready chatbots.
Facebook has posted a load of stats and research about its new state-of-the-art open source chatbot project called Blender (not to be confused with the open-source 3D graphics tool). The developers claim that Blender is “the largest-ever open-domain chatbot, outperforming others in terms of engagement and also feels more human, according to human evaluators.”
By others, it is aimed firmly at Google’s Meena, a bot getting ready to act as the front end on all sorts of Googly products. Google reckon that Meena can chat about anything, while Facebook claims theirs is better as it uses a new chatbot recipe that includes improved decoding techniques, novel blending of skills, and a “model with 9.4 billion parameters, which is 3.6x more than the largest existing system.” That’s compared to Google Meena’s measly 2.6 billion parameters. But you can bet that Google hasn’t been sitting still since its January reveal and can probably add a few billion more before these bots compete in the public arena.
Meena is billed as “an end-to-end, neural conversational model that learns to respond sensibly to a given conversational context.” while Facebook expects Blender to operate as the “the largest-ever open-domain chatbot.” In short, both will be able to talk to us like fellow adults, understanding the nuances and meaning of a conversation and digging out the appropriate responses.
Don’t believe the chatbot hype
All of this sounds exciting, and both products will undoubtedly wow the world with their public releases. They are likely to headline the next wave of Google hardware products, perhaps in Android 11 (or 12), where Google has already promised a “dedicated conversation section in the notification shade.” And Blender is likely to feature in Facebook’s next big social media platform or Messenger update as a better tool for businesses to communicate with customers.
But for most business or customer conversations they simply do not need this level of nuance or fluidity. The current level of AI chatbots can do the job perfectly well, which is why Facebook is already home to hundreds of thousands of business chatbots on Messenger, all helping businesses with the communications basics, customer support or product engagement.
For businesses with a wider footprint than Facebook, aka most of them, bots like SnatchBot provide the tools to design and launch a chatbot with no programming skills, to go on your website, app or other social media platform. It adds the natural language processing (NLP) power of AI to encourage deeper conversations with customers.
With a standard chatbot, most businesses can guide their customers from point A to point B, with a successful outcome and the required information they will have asked the bot for. All that without the need to talk about the weather, where people plan to go on holiday, or to analyse the conversation for deep nuance.
What currents bots can do is add multi-language support to broaden the audience, free up actual agents to deal with the more important or in-depth queries and provide constant availability in an increasingly 24/7 world. Bots can also help sell products, upsell accessories or take bookings and make appointments, all from within a friendly front-end that customers are rapidly getting used to.
At some point, the super chatbots of the future will be the norm, able to interact with other chatbots to do all of a customer’s digital bidding for them, but while the data and AI science-based articles (that sees Meena score 79% on the Sensibleness and Specificity Average (SSA), while Facebook doesn’t give its score, despite its apparent superiority) about these two future giants make for interesting reading, we won’t see them replacing current bot technology overnight. Indeed, when they do arrive, the complexity within them might put more businesses without expensive development teams off, even in the smart stuff is really just another service in the background.