Google's multimodal generative AI tool, Gemini, is poised for a comeback after its recent suspension due to image generation mishaps, according to Demis Hassabis, DeepMind's founder. Hassabis expressed confidence that the capability to produce images of people, temporarily halted last week, will be reinstated in the "next few weeks."
Speaking at the Mobile World Congress in Barcelona, Hassabis addressed concerns about Gemini's stumble, attributing the glitch to Google's oversight in discerning users' intentions. Rather than delving into technical intricacies, he emphasized the challenge of catering to diverse user prompts and the need for nuanced responses in advanced AI systems.
The incident underscores the complexities inherent in AI development, especially concerning historical accuracy and cultural sensitivity. Hassabis assured that Google is diligently rectifying the issue and aims to reintroduce the feature with enhanced precision in depicting historical figures.
In addressing the broader implications of generative AI, Hassabis acknowledged the multifaceted risks associated with its misuse by malicious actors. He stressed the imperative for collaborative efforts involving tech companies, governments, and civil society to establish ethical frameworks and safeguards against potential abuse.
Looking ahead, Hassabis envisioned a new era of AI-driven smart assistants poised to revolutionize daily interactions, potentially reshaping mobile device preferences. He hinted at the emergence of innovative form factors, such as smart glasses, to augment AI capabilities and enhance contextual understanding, paving the way for transformative advancements in technology.