Swiss keyboard startup Typewise has bagged a $1 million seed round to build out a typo-busting, ‘privacy-safe’ upcoming phrase prediction engine designed to operate entirely offline. No cloud connectivity, no details mining chance is the simple strategy.
They also intend the tech to function on textual content inputs produced on any system, be it a smartphone or desktop, a wearable, VR — or a thing weirder that Elon Musk may well want to plug into your mind in upcoming.
For now they’ve bought a smartphone keyboard application that is had all around 250,000 downloads — with some 65,000 active buyers at this stage.
The seed funding breaks down into $700K from extra than a dozen regional organization angels and $340K by means of the Swiss governing administration by means of a system (named “Innosuisse projects“), akin to a research grant, which is spending for the startup to utilize equipment studying professionals at Zurich’s ETH investigation university to build out the main AI.
The workforce soft introduced a smartphone keyboard application late very last calendar year, which features some extra tweaks (these as an optional honeycomb structure they tout as a lot more productive and the capability to edit following phrase predictions so the keyboard immediately groks your slang) to get customers to start feeding in information to establish out their AI.
Their main focus is on establishing an offline up coming term prediction engine which could be accredited for use any where end users are texting, not just on a mobile unit.
“The objective is to build a planet-leading text prediction motor that runs totally on-product,” says co-founder David Eberle. “The smartphone keyboard seriously is a 1st use situation. It’s fantastic to check and build our algorithms in a serious-lifestyle environment with tens of 1000’s of users. The much larger enjoy is to carry term/sentence completion to any software that entails text entry, on mobiles or desktop (or in long run also wearables/VR/Brain-Personal computer Interfaces).
“Currently it’s rather substantially only Google doing work on this (see Gmail’s car completion attribute). Apps these types of as Microsoft Teams, Slack, Telegram, or even SAP, Oracle, Salesforce would want such productiveness increase – and at that amount privateness/info protection matters a large amount. In the end we imagine that each and every “human-equipment interface” is, at the very least on the textual content-enter amount, driven by Typewise.”
You’d be forgiven for imagining all this appears a bit retro, offered the earlier boom in smartphone AI keyboards — these as SwiftKey (now owned by Microsoft).
The founders have also pushed certain features of their existing keyboard app — these kinds of as the exclusive honeycomb layout — prior to, likely down a crowdfunding route again in 2015, when they ended up calling the idea Wrio. But they reckon it’s now time to go all in — that’s why relaunching the small business as Typewise and shooting to develop a licensing business enterprise for offline upcoming word prediction.
“We’ll use the money to develop superior textual content predictions… to start with launching it in the keyboard application and then bringing it to the desktop to begin building partnerships with suitable software vendors,” says Eberle, noting they’re operating on a variety of enhancements to the keyboard application and also approach to devote on marketing to consider to strike 1M active customers future year.
“We have much more ‘innovative stuff’ [incoming] on the UX aspect as nicely, e.g. interacting with automobile correction (so the person can effortlessly intervene when it does something improper — in lots of international locations users just flip it off on all keyboards mainly because it will get irritating), gamifying the typical typing knowledge (significant opportunity for young ones/teens, also building them additional aware of what and how they sort), and so on.”
The competitive landscape around smartphone keyboard tech, mostly dominated by tech giants, has remaining area for indie performs, is the wondering. Nor is Typewise the only startup wondering that way (Fleksy has similar ambitions, for just one). On the other hand attaining traction vs such giants — and around very long established typing techniques — is the challenging little bit.
Android maker Google has ploughed resource into its Gboard AI keyboard — larding it with attributes. Even though, on iOS, Apple’s interface for switching to a third bash keyboard is infamously annoying and finicky the reverse of a seamless encounter. Additionally the indigenous keyboard delivers next word prediction baked in — and Apple has lots of privacy credit score. So why would a user bother switching is the issue there.
Competing for smartphone users’ fingers as an indie unquestionably is not straightforward. Different keyboard layouts and enter mechanism are generally a extremely tough sell as they disrupt people’s muscle memory and hit cell end users challenging in their convenience and productivity zone. Until the person is affected individual and/or stubborn adequate to adhere with a frustratingly unique expertise they’ll shortly ditch for the keyboard satan they know. (‘Qwerty’ is an historical typewriter structure turned typing habit we English speakers just simply cannot kick.)
Given all that, Typewise’s retooled concentrate on offline next term prediction to do white label b2b licensing helps make additional feeling — assuming they can pull off the main tech.
And, yet again, they are competing at a information downside on that entrance vs far more set up tech huge keyboard players, even as they argue that’s also a current market chance.
“Google and Microsoft (many thanks to the acquisition of SwiftKey) have a solid engineering in location and have began to present text predictions outside of the keyboard quite a few of their competitors, nevertheless, will want to embed a proprietary (challenging to make) or unbiased technological know-how, in particular if their value proposition is focused on privateness/confidentiality,” Eberle argues.
“Would Telegram want to use Google’s textual content predictions? Would SAP want that their clients’ facts goes through Microsoft’s prediction algorithms? Which is in which we see our suitable to acquire: environment-class textual content predictions that operate on-system (privateness) and are created in Switzerland (impartial surroundings, no security back doors, and so forth).”
Early impressions of Typewise’s subsequent phrase prediction smarts (gleaned by via checking out its iOS app) are rather reduced key (ha!). But it is v1 of the AI — and Eberle talks bullishly of getting “world class” builders doing work on it.
“The collaboration with ETH just started out a number of months in the past and as a result there are no significant enhancements yet noticeable in the dwell app,” he tells TechCrunch. “As the collaboration operates until the conclusion of 2021 (with the chance of extension) the huge vast majority of innovation is however to occur.”
He also tells us Typewise is doing work with ETH’s Prof. Thomas Hofmann (chair of the Facts Analytic Lab, formerly at Google), as well as possessing has two PhDs in NLP/ML and one particular MSc in ML contributing to the work.
“We get unique rights to the [ETH] know-how they really do not hold equity but they get paid out by the Swiss government on our behalf,” Eberle also notes.
Typewise suggests its smartphone application supports additional than 35 languages. But its subsequent term prediction AI can only take care of English, German, French, Italian and Spanish at this position. The startup claims a lot more are currently being additional.