Microsoft and Apple, the world's most valued companies, have recently been overtaken in value by Nvidia, whose silicon chip platform powers generative AI around the world. Nvidia now leads a select group of tech companies worth over 1 trillion US dollars – Microsoft, Apple, Alphabet and Amazon.
But there have been teething problems for companies marketing AI products - ranging from security issues to plagiarism lawsuits, and even to AI hallucinations (aromatic poisonous drinks anyone?)... Here are a few of the recent developments (and missteps).
Prior to Nvidia’s stockrise, Apple surged to the top of the stockmarket after launching Apple Intelligence in June – a personal intelligence system for iPhone, iPad and Mac. This will enable users to create AI-generated images in Keynote (its version of PowerPoint). They won’t be photorealistic images – instead, users will have three image options: animated images, illustrative images and sketch images.
Microsoft has added a dedicated Copilot key to Windows 11 PC’s. This provides quick access to the AI functions to the ChatGPT rival.
But the planned launch of the Recall feature, which takes a screenshot every 5 seconds, has been delayed because of controversy about security issues. The screenshots are set to be encrypted and stored locally with users requiring authentication to access – but there have already been demonstrations of how malware can be used to access the saved data and access sensitive information, such as bank details and passwords. So it’s back to testing, and finding the right balance between control and usability.
The Google Gemini chatbot (launched in December 2023) has had mixed success. As well as a confusing variety of releases – Nano, Pro and Ultra, and Gemini 1.5 Pro - and Gemini 1.5 Flash, there have been some teething problems.
One of these is the new Search summary (AI Overviews) which would be created at the top of your page. The AI powered results have included hallucinations in the results, for example: How many rocks shall I eat? (Image at right).
As these are quickly set loose online and end up as viral memes, Google has had to manually disable the AI summary for specific searches. These hallucinations have also occurred in ChatGPT – such as here in New Zealand where the Pak ‘n Save AI recipe generator produced this – an ammonia and bleach ‘perfect non-alcoholic’ drink.
Tech software company Adobe has also had some issues with its Terms of Service update. After a week of backlash from users, Adobe has had to clarify what their Terms of Service update does. Users noticed that the June update included a vague reference to “accessing content through automated and manual methods” and “techniques such as machine learning” leading to the assumption that Adobe was allowing itself to freely access and use customers’ work to train their generative AI models. The backlash went viral - “Adobe can not be trusted.” was a prominent meme and clarification had to be made by Adobe on June 18.
Note: Adobe has a Firefly AI model, which is trained on its own Adobe Stock images but has also copped complaints from artists (i.e. Beeple, Loish) whose work appears to have been used to train Firefly.
