This week in AI & Machine Learning: the ongoing conversation on AI and sentience, advancements in translation tech, and more.
A Note from the Author
The first weekend of summer is here and whether you’ll be hitting the beach, firing up the grill, or just relaxing, we hope you enjoy some fun in the sun. I’ll be starting my vacation this weekend, so look out for a guest post next week.
In other Plainsight news, our viral moment continues. This Precision Livestock Counting demonstration is still earning impressions in the tens of thousands every day. Check out our blog discussing the model and read our recent discussion of the ways computer vision can help agribusinesses and food processes avoid costly issues like the ongoing strawberry recall.
Artificial Intelligence News
Some Updates on the LaMDA Saga
In last week’s dispatch, we discussed Blake Lemoine, the Google engineer who has made headlines after publishing an interview with his employer’s advanced chatbot, LaMDA. In a televised interview with Tucker Carlson, Lemoine likened Google’s AI to a human child, adding, “any child has the potential to grow and be a bad person.” Lemoine has continued speaking to science and technology journalists to clear up lingering confusion. Previous reports have suggested that Lemoine hired a lawyer on LaMDA’s behalf. This week, however, Lemoine clarified that he only facilitated a meeting between the AI and the civil rights lawyer whose services it has decided to retain.
Google placed Lemoine on administrative leave last month and has roundly rejected the assertion that LaMDA is sentient. In general, expert consensus is that Google’s solution is far from developing a mind of its own, let alone a soul as Lemoine has suggested. The engineer’s statements nevertheless continue to drive conversations about what AI sentience would really mean and whether we have reason to fear it.
Meta’s Text-to-Speech Advancements
Translation software continues to evolve, but two significant roadblocks have historically stood in the way of progress. Speech-to-speech translation is typically a multi-step process with significant computational costs and 40% of the world’s languages lack written components altogether. Meta AI is at work on speech-to-speech translation (S2ST), a new approach that doesn’t require text. They believe their system, which breaks the translation process into discrete units, could lead to better translation systems in the future. The method has not only produced higher quality translations than traditional approaches, but proves more efficient with respect to metrics like runtime and max memory. Read the summary to learn more about Meta’s ongoing efforts and how S2ST amends and improves upon existing approaches to translation.
The Pentagon Offers Guidance on Responsible AI Use
The Department of Defense (DoD) first unveiled goals for ethical AI usage back in 2020. This week’s release of a document titled Responsible Pathway to AI Development and Acceleration represents the first efforts by the DoD to publicly outline steps for making responsable AI usage a reality. The lengthy document offers an overview of the Pentagon’s philosophy on AI as well as goals for how the DoD will embody the six core tenets of responsible AI, including traceability and governability. Securing trust is a key component of this mission. “Without trust,” the document reads, “warfighters and leaders will not employ AI effectively and the American people will not support the continued use and adoption of such technology.” Check out an overview of the document and interview with Deputy Secretary of Defense Kathleen Hicks.
Join our Community
See you next week! Until then, keep the conversation going on Plainsight’s AI Slack Channel.
About the Author & Plainsight
Bennett Glace is a B2B technology content writer and cinephile from Philadelphia. He helps Plainsight in its mission to make vision AI accessible to entire enterprise teams.
Plainsight’s vision AI platform streamlines and optimizes the full computer vision lifecycle. From data annotation through deployment, customers can quickly create and successfully operationalize their own vision AI applications to solve highly diverse business challenges.
