Superhuman: Shishir Mehrotra Faces AI Controversy
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Meeting with Shishir Mehrotra
Today, I had the opportunity to speak with Shishir Mehrotra, the CEO of Superhuman, which was previously known as Grammarly. The latter remains the flagship product of the company. Mehrotra, who also served as product director at YouTube and currently sits on the board of Spotify, is an influential figure in the tech industry. Initially, this interview, scheduled about a month in advance, was supposed to focus on the impact of artificial intelligence on software and creativity. However, the conversation took an unexpected turn.
Last August, Grammarly introduced a controversial feature called Expert Review. This feature allowed users to receive writing suggestions from "experts" cloned by AI. Journalists from The Verge and other media outlets discovered that their names, including mine, were used without permission. This revelation sparked outrage among journalists, notably Julia Angwin, who initiated a class action in response to the situation. In reaction, Superhuman first offered an email opt-out option before completely removing the feature. Shishir Mehrotra has apologized, and he will do so again during our interview.
During this time, I often wondered if Shishir would honor his commitment to record the episode of Decoder. My questions about decision-making and the impact of AI suddenly seemed more challenging to address. To his credit, he kept his promise and participated in the recording. The conversation was tense at times, revealing differing opinions on public perceptions of AI.
The Architecture of Superhuman
Superhuman positions itself as a productivity suite that integrates AI natively. The company changed its name from Grammarly to Superhuman to better reflect the expansion of its scope. In addition to Grammarly, which remains a highly sought-after writing assistant, Superhuman now offers Coda, a document space, and Mail, a popular email client.
A new product, Superhuman Go, has recently been launched. This platform aims to provide a proactive and personalized AI support network directly within users' work environments. For those familiar with Grammarly, Go is designed to enable anyone to create agents that function similarly. Whether for sales or support, these agents can assist users right where they work.
The central idea is that most AI tools require a significant change in behavior. Superhuman integrates AI where users are already working. Every day, around one million different applications and agents are used across their products, seamlessly incorporating AI into the user experience.
Decisions and Processes at Superhuman
Superhuman places great importance on informed decision-making. Shishir Mehrotra wrote an article titled Eigenquestions, which addresses the formulation of relevant questions to find the right solutions. Among the decision-making rituals, Dory and Pulse is a key process for gathering feedback and avoiding groupthink.
The Expert Review feature, which sparked controversy, offered text improvement suggestions based on expert advice, using names like journalists Casey Newton and Julia Angwin without their consent. Users were confused by the small checkmarks next to the names, suggesting a misleading officiality. In the face of criticism, Superhuman removed the feature even before there were any feedback or lawsuits. Shishir Mehrotra expressed that he was deeply saddened for not meeting the expectations of the experts, emphasizing that it was not a good feature and that it benefited neither the experts nor the users. It was quickly withdrawn, even before any criticism or legal actions arose.
In Conclusion
Superhuman employs around 1,500 people. The decision to launch the controversial feature was made by a small team, likely consisting of a product manager and a few engineers. Despite a structured decision-making process, the potential impact of unauthorized name usage was not anticipated. This raises questions about the vigilance needed in developing new features, especially when individual rights are at stake.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.