Are these people genuine or copies?
© Egypt Business Directory
Telefonica plans holograms
Telefónica is planning to introduce hologram telephony by 2026, whereby callers will be visible as 3D images. The technology uses VR glasses and is currently being tested in Munich. A joint project involving several providers, including Telekom, Vodafone and O2, is working on standards for cross-network holography. The required bandwidth is 20 megabits per second. O2 and Bitkom see great potential, especially in a professional context and for personal contacts, despite the need for additional devices such as VR glasses.
Deepfakes are getting better
The Alibaba Group has developed "Animate Anyone", an advanced technique for realistically animating people in photos. This new method shows significant improvements over previous approaches, but still has weaknesses with eyes and hands. The developers plan to release both demo and source code, which increases the potential for abusive deepfake applications.
WhatApp has avatars
WhatsApp now allows you to create an avatar based on a single photo. This global function allows subsequent personalisation of the avatar. Users can use their avatar as a profile picture or sticker and respond to status messages. Photo creation complements the manual method, whereby the photo is only used for this purpose and then deleted. However, early tests show that the function still has room for improvement.
Wombo has avatars
The new Wombo Me app from the creators of the Dream by Wombo AI art app allows users to create multiple lifelike avatars with just one selfie. Users can represent themselves as celebrities or film characters and try out different styles. The app offers a fast and user-friendly experience, with some avatar packs free and others paid.
ChatGPT fails Turing test
In an online Turing test study, the Eliza programme developed in 1966 convinced more people of its humanity than the free version of ChatGPT (GPT 3.5). Even the more advanced GPT 4.0 could not pass the Turing test. Human judges interacted with various "witnesses", including humans and AI models. Eliza achieved a success rate of 27 per cent, which was higher than that of all GPT 3.5 versions and some GPT 4 variants.