Technology and Public Purpose Project

PHOTO RECAP: Thanks to all who attended #TAPPSummit23 - we had a great time engaging in discussion on the cutting edge of what's happening in technology and public purpose and featuring the important research work of our fellows!

Thank you to everyone who has contributed to the mission of TAPP and supported our work throughout our time at Harvard! Our research and events will continue to exist on the Belfer Center’s website here: buff.ly/3QCejMn #CarrytheTorch

Genome editing therapies are in the early stage of trials in humans, so the full extent of feasibility, safety, and impacts are still unknown. Public purpose considerations include technical feasibility, safety, and ownership guidance, among others. Read more in our factsheet!

Genome editing is not new—scientists have used various techniques to create genetically modified organisms (GMOs) for decades. Yet the application to humans was limited because the methods were slow, costly, and not considered safe or reliable to deliver the treatment in humans.

Check out our factsheet on genome editing! 🧬 Genome Editing (also called “gene editing”) is a group of technologies that give scientists the ability to change an organism’s DNA. buff.ly/44wPANr

In April, China published a draft regulation on generative AI. It has a mix of sensible restrictions on AI risks + signals its typically aggressive intervention in tech. More here: buff.ly/3KAV3Lp

For enterprises deploying AI responsibly and efficiently, enterprises use best practices like ModelOps to ensure these deployments continuously meet regulatory requirements. Check out this podcast to learn more about AI governance essentials here: buff.ly/3DkFK5k

Former OpenAI founders built Anthropic to create a safe AI company. Their experiments involve asking a powerful LLM they've created to deceive them, in order to figure out how to prevent this behavior. buff.ly/43JzizL

OpenAI marshaled a global pipeline of human labor for 2+ years to enable ChatGPT & GPT-4 to exist. Workers who dealt with the worst parts are still suffering PTSD more than a year later. My story with @dseetharaman out today on the front page of @WSJ. wsj.com/articles/chatg…

2 spyware companies were added to a federal blacklist preventing US companies from engaging in certain trade activities with them, part of an effort to address the proliferation and misuse of commercial spyware: buff.ly/3Oua2ZF

Silicon Valley start-up Cerebras revealed its newest supercomputer last week, built with specialized chips designed to power AI products. More AI supercomputers might be on the horizon: buff.ly/44PXVeU

Last week, Meta unveiled LLaMa 2, its first LLM available to anyone to use for free. It's part of Meta's new suite of AI models, giving people options for closed-source or open-source approaches for their projects. technologyreview.com/2023/07/18/107…

“Having access to the experimental results that platforms themselves use to understand their products is a simple, powerful step toward effective transparency.” - @NateLubin & @RaviDataIyer for @lawfare lawfaremedia.org/article/how-te…

While differential privacy is increasingly leveraged to protect sensitive data against potential privacy attacks, policy is falling short. The privacy framework must be updated to promote investment & adoption of differential privacy tools. Read more in our primer!

Also, since there's no standards per se, companies are currently using differential privacy with weaker privacy protection than the research community finds acceptable.

FACTSHEET FRIDAY: Differential privacy is a safeguard used to protect an individual’s data privacy. It's a criterion that allows for the collection & publication of data patterns & trends, while protecting the privacy of individuals captured in a dataset. buff.ly/2OUFOme

A UN group set a goal for net-zero emissions for global shipping by ~2050, which is ambitious for an industry that's tough to decarbonize - but it may yet be accomplished. Slowing down ships and using biofuels are some potential routes. buff.ly/3Q0FIac

What is generative AI's impact on worker productivity? A case study from Stanford @DigEconLab and @MIT suggests access to AI assistance increased productivity by 14%, with the biggest impact on less experienced workers. stanford.io/3DrbG7W

The new EU/US Data Privacy Framework allows EU citizens to object when they believe their personal info has been collected improperly by US intelligence agencies - resolving an ongoing struggle btn US national security concerns & European privacy rights. buff.ly/3JQPiZo

The startup Writer builds AI tools that produce content in the voice of a specific person or brand. They imagine their tools helping a smaller # of humans writers to more efficiently and cost-effectively produce the work of many. buff.ly/43nBFaW