13 - Technology Whisperers Season 2 Launch episode
Manage episode 374232706 series 3303307
Contact Details
Alistair Ross
alistair@revolutioninfosec.com
Web: https://revolutioninfosec.com
Linkedin: https://www.linkedin.com/in/alistairjross https://www.linkedin.com/company/revolutioninfosec
Sean G Muller
seangmuller@technologyleader.co.nz
Linkedin: https://www.linkedin.com/in/sgmuller/
Show Notes:
1) OpenAI are being sued.
Since the inception of ChatGPT back in November 2021, controversy has never been far away, and smaller legal battles have always been looming, however, at the end of June, A firm in California filed a class-action lawsuit for "stealing" personal data in order to train ChatGPT.
What is interesting about this case is that depending upon how well prepared the complainants are, OpenAI may be required to fess up to how they trained their model. The only public information about OpenAI's ChatGPT that we really know is that OpenAI scraped 300 billion words from the Internet. We don't know how they scraped them and where they scraped it from. The complainant states that OpenAI trained its model partially with personal information, some of it from social media accounts, including Twitter and Reddit. The complainant goes on to state that OpenAI did so in secret, without registering as a data broker as was required to do under applicable laws.
The most interesting aspect for me is by how much OpenAI are profiting from these sources, or whether they in fact are profiting, OpenAI have oft stated that they do not (yet) turn a profit. Certainly, on paper, OpenAI are worth many billions in unrealised value, and recently Microsoft made a $10B USD stake in them, so that might have turned them into a profit making machine. The fifteen counts in the complaint include privacy violations, negligence in failing to protect personal information, as well as larceny by obtaining large amounts of personal data to train its models by illegal means. The lawsuit has stretched as far as to seek injunctive relief in the form of a temporary freeze on commercial use of OpenAI's products. If granted, this could have vast reaching effects on organisations that currently use OpenAI's model on their own applications or services.
Where do you sit on this Sean? It could be argued that the information that is publicly accessible on the Internet is free for anyone to see, including OpenAI, regardless of whether that is ethically responsible, it certainly is 'freely available'. On the other hand, it could be argued that people have a right to be guardians of their own content. If it is to be used in a wholesale manner by another corporation (potentially for profit), then they must seek permission from the original content author. Doing otherwise could be seen as a violation of privacy.
2) Elon Musk launches new AI company called xAI.
A team around of 12 engineers. His mission he states to 'Understand the true nature of the Universe'. On the website, it states that more information is coming in the following weeks and months, so I won't expect to see anything like an OpenAI or Bard competitor for the foreseeable future. However, they certainly have the chops to do it, remembering that Musk was one of the original founders of OpenAI.
What is interesting is that Musk was also one of the people who recently spoke out about AI as a whole, and took issue with the rapid pace of development (mainly pointed at OpenAI), stating that we should press the pause button on AI development. By launching this startup, it's clear that he's made a u-turn on that decision!
The website states that it is a standalone company, unrelated from Tesla and other Musk companies including Twitter (now called X corp), however Musk states that he will work closely with them to make progress towards xAI's mission. From a privacy perspective, will this mean that it could use tweets (including DMs) as a training model for xAI? Obviously this is possibly where OpenAI are in hot water, as per my previous story, so it will be interesting to find out how transparent Musk will be about the training and development of xAI. However, Musk is well known to be critical about the less than transparent nature of OpenAI, he would likely have a lot of egg on face if he too, chose to be less than open about the training of his model.Obviously this is possibly where OpenAI are in hot water, as per my previous story, so it will be interesting to find out how transparent Musk will be about the training and development of xAI. However, Musk is well known to be critical about the less than transparent nature of OpenAI, he would likely have a lot of egg on face if he too, chose to be less than open about the training of his model.Obviously this is possibly where OpenAI are in hot water, as per my previous story, so it will be interesting to find out how transparent Musk will be about the training and development of xAI. However, Musk is well known to be critical about the less than transparent nature of OpenAI, he would likely have a lot of egg on face if he too, chose to be less than open about the training of his model.Obviously this is possibly where OpenAI are in hot water, as per my previous story, so it will be interesting to find out how transparent Musk will be about the training and development of xAI. However, Musk is well known to be critical about the less than transparent nature of OpenAI, he would likely have a lot of egg on face if he too, chose to be less than open about the training of his model.
3) China has taken a major step in regulating generative AI like ChatGPT
this happened a couple of weeks ago, making it one of the first countries in the world to do so. The first draft of the law was fairly substantial in its punitive measures, which would see companies and individuals breaking the law fined up to 100,000 yuan. The wording was relaxed to allow development of secure and trustworthy chips, software, tools and computing power, however the key provisions in the new law which is to go into law on the 15th of August will require AI service providers to conduct routine, thorough security reviews and all of their algorithms must be registered with the Beijing government. China wants to press ahead in the world of AI, but has lacked the freedom to do so from its government. Companies like Baidu and Alibaba have not released any mainstream AI products (probably for fear of reprisal from the government). It will be interesting to see if this stifles innovation or allows China to leapfrog the USA. Eyes will be on the China following 16th August to see what Alibaba, JD and Baidu have been developing in the background.
4) Bard's quiet updates. On the 14th July, Google quietly released an update to Bard.
The groundbreaking update to Bard is the ability for it to accept images in its prompts. You can take a photo of your fridge and ask Bard to give you some recipes to rustle up, for example.
5) In other update news, ChatGPT got a new Beta feature that is called "Code Interpreter".
With this, you can give ChatGPT a prompt and with the interpreter switched on, it will know how to write and execute python code. It can also accept with file uploads. This is extremely helpful for aspects of data analysis, image conversions, or editing a code file. OpenAI say that the data does not persist past a session and is deleted after the session is gone.
29 jaksoa