Swiss keyboard startup Typewise has bagged a $1 million seed spherical to construct out a typo-busting, ‘privacy-safe’ following phrase prediction motor created to operate fully offline. No cloud connectivity, no details mining danger is the simple strategy.
They also intend the tech to perform on text inputs made on any gadget, be it a smartphone or desktop, a wearable, VR — or something weirder that Elon Musk may want to plug into your mind in future.
For now they’ve bought a smartphone keyboard application which is experienced all-around 250,000 downloads — with some 65,000 lively consumers at this issue.
The seed funding breaks down into $700K from far more than a dozen neighborhood business enterprise angels and $340K by way of the Swiss federal government as a result of a system (referred to as “Innosuisse projects“), akin to a exploration grant, which is paying for the startup to use machine finding out experts at Zurich’s ETH investigate college to construct out the core AI.
The team smooth launched a smartphone keyboard app late past year, which incorporates some supplemental tweaks (these kinds of as an optional honeycomb format they tout as far more productive and the means to edit upcoming word predictions so the keyboard rapidly groks your slang) to get end users to get started feeding in details to create out their AI.
Their most important concentrate is on building an offline future term prediction engine which could be certified for use any where people are texting, not just on a cell machine.
“The target is to produce a planet-major textual content prediction engine that operates fully on-device,” says co-founder David Eberle. “The smartphone keyboard definitely is a initially use case. It’s great to test and produce our algorithms in a true-daily life location with tens of thousands of consumers. The greater engage in is to bring phrase/sentence completion to any software that includes text entry, on mobiles or desktop (or in potential also wearables/VR/Brain-Personal computer Interfaces).
“Currently it’s rather considerably only Google doing the job on this (see Gmail’s automobile completion feature). Apps this sort of as Microsoft Teams, Slack, Telegram, or even SAP, Oracle, Salesforce would want these kinds of efficiency raise – and at that amount privacy/facts security matters a whole lot. Eventually we visualize that every “human-machine interface” is, at least on the textual content-enter amount, powered by Typewise.”
You’d be forgiven for considering all this sounds a bit retro, offered the previously boom in smartphone AI keyboards — these as SwiftKey (now owned by Microsoft).
The founders have also pushed precise aspects of their present-day keyboard app — these types of as the exclusive honeycomb format — just before, heading down a crowdfunding route again in 2015, when they ended up calling the strategy Wrio. But they reckon it’s now time to go all in — that’s why relaunching the business as Typewise and capturing to establish a licensing organization for offline future term prediction.
“We’ll use the resources to establish sophisticated text predictions… very first launching it in the keyboard application and then bringing it to the desktop to commence constructing partnerships with related computer software vendors,” suggests Eberle, noting they’re operating on numerous enhancements to the keyboard app and also approach to invest on internet marketing to try to strike 1M energetic customers future year.
“We have extra ‘innovative stuff’ [incoming] on the UX side as very well, e.g. interacting with vehicle correction (so the consumer can simply intervene when it does something erroneous — in several countries consumers just switch it off on all keyboards for the reason that it receives irritating), gamifying the common typing encounter (big possibility for young ones/teens, also building them additional mindful of what and how they kind), and many others.”
The aggressive landscape all-around smartphone keyboard tech, largely dominated by tech giants, has still left home for indie plays, is the considering. Nor is Typewise the only startup contemplating that way (Fleksy has equivalent ambitions, for just one). On the other hand gaining traction vs such giants — and around extensive recognized typing techniques — is the tricky bit.
Android maker Google has ploughed resource into its Gboard AI keyboard — larding it with capabilities. Though, on iOS, Apple’s interface for switching to a 3rd party keyboard is infamously irritating and finicky the reverse of a seamless experience. In addition the indigenous keyboard features upcoming phrase prediction baked in — and Apple has lots of privateness credit score. So why would a consumer bother switching is the dilemma there.
Competing for smartphone users’ fingers as an indie unquestionably is not straightforward. Substitute keyboard layouts and enter system are constantly a really tricky offer as they disrupt people’s muscle mass memory and hit cellular end users really hard in their convenience and productivity zone. Except if the person is affected person and/or stubborn adequate to adhere with a frustratingly unique expertise they’ll shortly ditch for the keyboard satan they know. (‘Qwerty’ is an ancient typewriter format turned typing routine we English speakers just simply cannot kick.)
Supplied all that, Typewise’s retooled target on offline next word prediction to do white label b2b licensing tends to make much more sense — assuming they can pull off the main tech.
And, again, they are competing at a data drawback on that entrance vs much more founded tech big keyboard players, even as they argue that is also a marketplace chance.
“Google and Microsoft (many thanks to the acquisition of SwiftKey) have a sound technology in area and have started out to offer you text predictions outside the house of the keyboard many of their competitors, having said that, will want to embed a proprietary (tricky to establish) or unbiased technology, in particular if their price proposition is focused on privateness/confidentiality,” Eberle argues.
“Would Telegram want to use Google’s textual content predictions? Would SAP want that their clients’ information goes by means of Microsoft’s prediction algorithms? That is in which we see our suitable to acquire: globe-class text predictions that operate on-unit (privacy) and are created in Switzerland (impartial atmosphere, no security again doorways, and so on).”
Early impressions of Typewise’s subsequent phrase prediction smarts (gleaned by by means of checking out its iOS application) are fairly reduced vital (ha!). But it’s v1 of the AI — and Eberle talks bullishly of acquiring “world class” developers working on it.
“The collaboration with ETH just started off a couple weeks in the past and therefore there are no sizeable improvements nevertheless visible in the live application,” he tells TechCrunch. “As the collaboration operates until the finish of 2021 (with the prospect of extension) the vast the vast majority of innovation is however to arrive.”
He also tells us Typewise is operating with ETH’s Prof. Thomas Hofmann (chair of the Facts Analytic Lab, formerly at Google), as properly as obtaining has two PhDs in NLP/ML and 1 MSc in ML contributing to the energy.
“We get exclusive legal rights to the [ETH] engineering they do not hold equity but they get paid by the Swiss federal government on our behalf,” Eberle also notes.
Typewise claims its smartphone app supports much more than 35 languages. But its up coming term prediction AI can only take care of English, German, French, Italian and Spanish at this place. The startup says much more are becoming extra.