Chat GPT’s release at the end of 2022 has, so far, been the dominating tech story of 2023. For many people outside the tech field, this is the first time they’ve been aware of just how similar to human interaction an AI tool can be, and it’s both thrilling and terrifying to imagine what will come next. At DataKind, we’ve been following the build up of generative AI tools, and are eager to explore the opportunities that these could bring to the global social sector.
“There’s been a huge influx of activity in generative AI, some of these – like ChatGPT, DALL-E2 for image generation, and Stable Diffusion have been become household names, but this is just the tip of the iceberg,” says Matthew Harris, Head of Data Science at DataKind. “There are many other models being developed across a wide range of domains, and Google and Microsoft are racing to include generative AI in their products.”
Don’t Believe (all) the Hype (Yet)
This week, DataKind held a webinar explaining and reflecting on the underlying technology of these tools, generative AI, and the implications for the nonprofit, government, and other direct service providers who partner with DataKind to scale their reach and impact using AI. The upshot is that there are so many ways in which this ability to “sound human” and make connections on its own could enable massive scaling and greater efficiency of everything from social services and resource distribution to better and more accessible digital records; but only if the tech sector is successful in mitigating it’s risks and concerns.
These concerns include copyright violations and environmental impact of the massive data centers required to run these tools, as well as the potential to amplify the already present deep-fakes and bias already rampant on social media. There’s the issue of transparency, as these models are commercial and privately-owned, and there’s a lack of oversight. There’s also one of the most vexing problems for developers: hallucination.
“Large language models can sound so convincing, and as human beings we’re conditioned to develop trust if something responds to us in a human-like fashion,” says Harris, “so it makes it very dangerous when they get things wrong.”
There are, however, many reasons to be optimistic, or at least to reserve judgment for now. We’re still very early in development of these products, and there are massive resources being poured into managing these concerns. Despite the record-levels of daily active users already on Chat GPT, it’s just a pilot and not yet available as a commercial product. Microsoft, Google, and Meta have all made substantial investments in the technology with plans to integrate it with their current products and develop new ones. There are responses to the closed commercial models, such as the open-source and fully transparent Large Language Model BLOOM.
It’s encouraging that these big companies are investing in many of the guardrails and infrastructure that would ease these concerns. In addition, the high visibility of this technology brings a high level of public scrutiny, and tech companies are well aware that if they want consumers to embrace these products, AI safety must be a priority. If, in particular, the safety and hallucination issues are not solved, generative AI may become another overly-hyped technology that fails to deliver in real-world scenarios.
How Can Your Organization Use Generative AI?
DataKind will continue to track the development of generative AI, and will hold additional webinars and create both technical and non-technical resources for the social sector as we see new opportunities emerge. Curious how your organization can use AI? Follow our webinar series so you can up skill your organization. We’ll soon be releasing information on our next Nonprofit Data 101 webinar. Stay tuned! For now, participants praised both the fairness and accessibility of this webinar, with the right balance of technical detail and context for non-technical audiences.
“This was an outstanding overview of generative AI as it relates to the social sector. I’ve raved about it to everyone!”
“I found the session really informative and interesting. The speakers had so many interesting and useful things to say, there was no ‘filler’ at all!”
“Best presentation I’ve seen for the social sector and small nonprofits. Balanced and actionable!”
Join the DataKind movement.
- Interested in supporting our work? Donate here.
- Interested in sponsoring a project? Partner with us.
- Interested in volunteering with DataKind? Look no further.
- Interested in working at DataKind? We’re hiring!
- Interested in submitting a project? Go for it!