By NewsBlogger California has introduced guidelines for the use of generative AI technologies, requiring audits and risk assessments before adoption, along with specific guidance on purchasing, training, and inventory submission, while OpenAI faces criticism for training AI models without copyrighted materials.
At a glance
- California introduces guidelines on generative AI technologies, requiring audits and risk assessments before adoption.
- Providers must go through a written solicitation process to offer services, and state agencies must involve generative AI experts in drafting contracts.
- Agency CIOs must conduct risk assessments of generative AI technologies before procurement using specific principles.
- State entities must submit an inventory of generative AI uses to the California Department of Technology.
- KL3M, a new Language Model, challenges assumptions about training AI models with copyrighted materials and has received a “Licensed Model (L) Certification.”
The details
California has recently introduced guidelines on the use of generative AI technologies, requiring audits and risk assessments before adopting new systems.
These guidelines follow Governor Gavin Newsom’s executive order from September 2023. State agencies are now mandated to demonstrate the generative AI aspects of products or services offered by companies.
The guidelines also offer specific guidance on both “incidental” and “intentional” purchases of generative AI technologies.
Providers must go through a written solicitation process to offer services, while state agencies must bring in a generative AI expert to help draft and check contracts for new systems.
Agency CIOs are required to conduct risk assessments of generative AI technologies before procurement, using the California Generative Artificial Intelligence Risk Management Principles.
Workforce training is also mandated before adopting generative AI technologies, with executive-level staff receiving initial training followed by legal and privacy specialists, and general staff receiving training before generative AI tools are deployed.
Additionally, all employees must undergo information privacy and security training.
State entities are further required to submit an inventory of generative AI uses to the California Department of Technology, and agencies have been supplied with a toolkit written in plain language for non-technical staff to help procure generative AI technologies.
This toolkit will be continuously updated as best practices evolve.
OpenAI has stated that training leading AI models without using copyrighted materials would be impossible.
Critics, however, argue that OpenAI should have sought consent or paid licensing fees for copyrighted data.
A new Language Model (LLM) called KL3M challenges this assumption, as it was trained on public domain data collected by 273 Ventures, co-founded by Daniel Martin Katz and Michael Bommarito.
KL3M received a “Licensed Model (L) Certification” from Fairly Trained, which is awarded to companies that prove their AI model training data was obtained legally.
KL3M is the first LLM to receive this certification and has two versions: kl3m-170m and kl3m-1.7b, with 273 Ventures preparing to release a 3.7-billion parameter variant of KL3M. Advertised as helpful for drafting legal documents and beyond, KL3M generalizes well beyond the legal sector and has scored lower on toxic outputs compared to other models.
KL3M is already in use among law-firm customers, although its cost is not publicly available.
Motorola and Samsung have showcased flexible smartphones that can double as smartwatches at the Mobile World Congress.
Deutsche Telekom revealed a handset that replaces standard apps with an AI assistant, known as the ‘T-phone’ device, which reimagines user interaction with mobile devices.
According to Deutsche Telekom executives, traditional apps may become obsolete within the next decade, indicating a potential shift in the industry.
Smartphone sales have been declining in recent years, with global shipments peaking in 2016, leading analysts to believe that the world has passed “peak smartphone” and manufacturers are now seeking to reinvent the industry.
Recent product launches have shown potential directions for the post-smartphone era, with AI startup Rabbit unveiling the standalone R1 device at the Consumer Electronics Show.
Leading phone makers are expected to continue launching upgrades to their smartphone ranges, while generative AI is predicted to impact how smartphones are used in the future.
Qualcomm’s CEO also believes that artificial intelligence will fundamentally alter the relationship people have with their phones, with generative AI potentially becoming the interface between humans and applications in the future.
Article X-ray
Sources
Here are all the sources used to create this article:
Facts attribution
This section links each of the article’s facts back to its original source.
If you suspect false information in the article, you can use this section to investigate where it came from.
| aibusiness.com |
|---|
| – California has introduced guidelines on the use of generative AI technologies – Guidelines require audits and risk assessments before adopting new systems – Guidelines follow Governor Gavin Newsom’s executive order from September 2023 – State agencies must show generative AI aspects of products or services offered by companies – Guidelines offer specific guidance on “incidental” and “intentional” purchases of generative AI technologies – Providers must undergo a written solicitation process to offer services – State agencies must bring in a generative AI expert to help draft and check contracts for new systems – Agency CIOs must conduct risk assessments of generative AI technologies before procurement – Risk assessments must use California Generative Artificial Intelligence Risk Management Principles – Workforce training is required before adopting generative AI technologies – Executive-level staff will receive initial training followed by legal and privacy specialists – General staff will receive training before generative AI tools are deployed – All employees will need to take information privacy and security training – State entities must submit an inventory of generative AI uses to the California Department of Technology – Agencies have been supplied with a toolkit to help procure generative AI technologies – Toolkit is written in plain language for non-technical staff and will be continuously updated as best practices evolve |
| venturebeat.com |
|---|
| – OpenAI stated that it would be impossible to train leading AI models without using copyrighted materials – Critics argue that OpenAI should have sought consent or paid licensing fees for copyrighted data – A new LLM called KL3M challenges the assumption that copyrighted data is necessary for creating a useful model – KL3M is the work of 273 Ventures, co-founded by Daniel Martin Katz and Michael Bommarito – KL3M received a “Licensed Model (L) Certification” from Fairly Trained – Fairly Trained certification is awarded to companies that prove their AI model training data was obtained legally – KL3M was the first LLM to receive this certification – KL3M was trained on public domain data collected by 273 Ventures – KL3M has two versions: kl3m-170m and kl3m-1.7b – 273 Ventures is preparing to release a 3.7-billion parameter variant of KL3M – KL3M is advertised as helpful for drafting legal documents and beyond – KL3M generalizes well beyond the legal sector – KL3M’s performance was benchmarked against other models in its class – KL3M scored lower on toxic outputs compared to other models – KL3M is already in use among law-firm customers – |
| The cost of KL3M is not publicly available |
| independent.co.uk |
|---|
| – Motorola and Samsung showcased flexible smartphones that can double as smartwatches at Mobile World Congress – Deutsche Telekom revealed a handset that replaces standard apps with an AI assistant – The ‘T-phone’ device reimagines user interaction with mobile devices – Traditional apps may become obsolete within the next decade, according to Deutsche Telekom executives – Smartphone sales have declined in recent years, with global shipments peaking in 2016 – Some analysts believe the world has passed “peak smartphone” and manufacturers are looking to reinvent the industry – Recent product launches have shown potential directions for the post-smartphone era – AI startup Rabbit unveiled the standalone R1 device at the Consumer Electronics Show – Leading phone makers are expected to continue launching upgrades to their smartphone ranges – Generative AI is expected to impact how smartphones are used in the future – Qualcomm’s CEO believes artificial intelligence will fundamentally alter the relationship people have with their phones – Generative AI may become the interface between humans and applications in the future |
