Confluence for 10.1.23
Confluence for 10.1.23
A release date for Microsoft 365 Copilot. The implications of skill leveling. Universal language translation. Developments on the legal front. OpenAI's big week and the multi-modal turn.
Welcome back to Confluence. We’re almost two months into sharing ideas and information on the intersection of AI and corporate communication here, and it’s not lost on us how much has changed just in these six weeks. For some time we’ve been saying that a gulf exists between what most people know about generative AI and what it means for them. If anything, the past six weeks has us believing that gulf is widening as the technology evolves faster than many peoples’ experience with it.
This was brought into sharp relief for us last week. Several of our colleagues were leading a session on generative AI for a group of clients, taking them through a high-level orientation to the technology and its strengths and weaknesses specific to corporate communication. Several hours of the program engaged those clients in hands-on use cases for GPT-4. For some it was their first experience using the technology for anything more meaningful than something amusing or perhaps answering simple questions, and for many it was their first exposure to a large language model of any kind. At several points in the session we noted the sound of audible gasps in the room — gasps of surprise, amazement, and in some cases, recognition.
We don’t believe we are overstating things when we claim that a significant transformation of corporate communication process and practice is at our doorstep. It will have profound implications for what we do and how we do it, with deep second-order consequences in people, process, content, and tool sets. But unlike other waves of technological transformation, like email, or the internet, the pace of this transition will be much faster. To paraphrase Alvin Toffler, it’s better to ride a wave than to fight it. If you’ve been reading this newsletter but not getting your hands dirty with the tools, and more important, the thinking and ideas about their consequences, it’s time to start. We believe generative AI is going to unlock incredible potential and value for communication professionals, but that unlocking will still mean significant change — and the more you know about that change, the easier it will be for you to manage.
That said, it was another big week in AI developments. Here’s what has our attention at the intersection of AI and corporate communication as we open October, 2023:
Microsoft Announces November 1 Release Date for 365 Copilot
The Implications of Skill Leveling
Universal Language Translation
Developments on the Legal Front
OpenAI’s Big Week and the Multi-Modal Turn
Microsoft Announces November 1 Release Date for 365 Copilot
One of the most anticipated developments in AI is officially one month away.
Since Microsoft initially announced plans for 365 Copilot earlier this year, we’ve been advising clients that its arrival would be a major inflection point for the adoption and proliferation of generative AI in organizations. You may see a somewhat dated Microsoft demo video here. Last week, amid a flurry of other announcements, Microsoft revealed the much-awaited general release date for 365 Copilot: November 1.
At that point, the ball will be in organizations’ courts. This is not a “flip the switch” moment where Copilot will appear for all Microsoft Office users. Organizations will still need to decide whether to pay the $30 per user per month, determine implementation approaches and timelines, and all of the other steps that come with enterprise software. But on November 1, the technology will be available for any organization that wants it. Beyond November 1, the things to watch will be 1) how many organizations do buy it, 2) what that does for generative AI adoption across those organizations, and 3) the second- and third-order consequences of that adoption. We have an internal team ready to be early adopters as soon as we’re able, and will share our first-hand experiences here.
The Implications of Skill Leveling
Ethan Mollick posts an interesting embellishment on the implications of generative AI for professionals.
Ethan Mollick recently posted an article on the implications of AI acting as a “skill leveler,” one of the more notable findings from the recent Harvard working paper on the effects of generative AI on knowledge worker productivity and quality. We suggest reading Ethan’s post in full, and it reiterates the important emerging finding that AI acts as a skill leveler for a huge range of professional work — making the worst performers in an organization much better. The part of the post that really got our attention, though, was Ethan’s thinking on what this skill leveling may mean for the distribution of skills in a team, organization, or market:
Just because early results for AI suggest that only lower performing people benefit does not mean that this is the only possible pattern. It may be that the reason only lower performers gain from AI currently is because the current AI systems are not good enough to help top performers. Or, alternately, it might be that top performers need more training and work to get benefits from AI. If either of these conditions prove true, and they certainly seem plausible, then AI might act more as an escalator, increasing the skills for everyone, from top to bottom performers. After an adjustment period, the relative skill positions stay similar, but everyone gets more done, faster.
Alternately, it might be that some people are just really good at working with AI. They can adopt Cyborg practices better than others and have a natural (or learned) gift for working with LLM systems. For them, AI is a huge blessing that changes their place in work and society. Other people may get a small gain from these systems, but these new Kings and Queens of AI get orders of magnitude improvements. If this scenario is true, they would be the new stars of our AI age, and are sought out by every company and institution, the way other top performers are recruited today.
We don’t know the ultimate shape of the new post-AI skills distribution, but we do absolutely know that things are changing. Even with the relatively primitive tools of our current, unspecialized AI systems, it is clear that we can become much more productive, and that less-skilled workers are now at much less of a disadvantage than they used to be.
Many people in corporate communication are focusing and trying to define how generative AI can become an asset to their teams and the work they do, and rightly so. That said, we are also working very hard to understand and anticipate the second-order consequences, both positive and negative. The effects they will have on talent distribution is clearly one of these spaces, and it’s one to keep on your radar as things develop.
Universal Language Translation for Leadership Communication
A future where leadership messages are translated in leaders’ own voices.
Translation, for communication teams, has historically involved tradeoffs across cost, timeliness, languages, and more. With developments in AI, many of those constraints are disappearing, seemingly by the day. We already saw that large language models like GPT-4 are capable of translating text across a number of languages, quickly and (mostly) accurately1. That appears to be just the beginning, with even more compelling advances in AI audio and video pointing to a future where the audio of leaders’ messages can be translated into multiple languages, in their own voice.
Spotify’s Voice Translation for podcasts is a preview into what that future could look like, at least for audio. With this feature, users can now listen to a number of popular podcasters like Lex Fridman and Dax Shepard in French, German, Spanish and a number of other languages — in the podcaster’s own voice. This is not a translator restating the original text, it’s AI doing the translation and then producing the audio in the speaker’s original voice. And it’s uncanny in its quality.
What does this mean for corporate communication? A new opportunity for conveying leadership communication. With generative AI, we expect communication teams will have the ability to produce town halls, videos, and podcasts in which leaders communicate with employees in native languages, in their own voices. And eventually, we presume the technology will allow this translation to occur in real time. This certainly appeals to us, and we expect it will to you as well.
Developments on the Legal Front
As lawsuits over intellectual property intensify, indemnification emerges as a “feature” across AI products.
A lawsuit against Open AI by prominent authors including George R.R. Martin and John Grisham is the latest in a series of similar claims by artists against the companies behind some of the biggest AI models. The issue at the center of each case is whether these companies improperly used the artists’ work in training their models. Legal experts are divided on what judgments we should expect, but regardless of where these cases land, they will set important precedents for the future of these models.
With that legal uncertainty as backdrop, more companies are including indemnification as part of their product offerings in an attempt to assuage the concerns of wary customers. Adobe has included an indemnification clause with its Firefly product, while Getty Images has launched its own generative AI product trained exclusively on its own library of licensed content.
It’s likely that we’ll be operating in a legal grey area for the foreseeable future as these early lawsuits make their way through the legal system. As Adobe’s indemnification clause and Getty’s licensed image generator show, however, the benefits of these technological advances are going to be available for even the most risk-averse organizations.
OpenAI’s Big Week and the Multi-Modal Turn
New developments will — very soon — transform how we use generative AI.
As of today, for an increasing number of Plus users GPT-4 can now or will soon be able to listen, talk, see, and create images. With these developments, prompt engineering looks to be becoming less important — users can use natural language to develop and edit high-quality imagery, and can upload images as a way of conveying context to the model in instances where an image is worth a thousand words.
All this is working to reduce the friction associated with engaging with and deriving benefit from generative AI. As this friction drops, relative utility and exposure should increase, as should the pace of adoption. The bottom line: more people using the technology in their daily lives, with greater ease, and bringing expectations for similar ease and utility to how they use (or don’t use) these tools at work.
That’s all we have for this edition of Confluence. We’ll leave you with GPT-4V(ision) finding Waldo.
AI Disclosure: We used generative AI in creating imagery for this post. We also use it selectively as a summarizer of content and as an editor and proofreader.As an example, here is a copy-and-paste of GPT-4’s translation of this section of Confluence into Spanish. We gave no specific prompting on voice or style. Those who speak Spanish can judge the quality:
La traducción, para los equipos de comunicación, históricamente ha implicado concesiones entre coste, puntualidad, idiomas y más. Con los avances en la IA, muchas de esas limitaciones están desapareciendo, aparentemente día a día. Ya vimos que modelos de lenguaje grandes como GPT-4 son capaces de traducir texto en una variedad de idiomas, rápidamente y (en su mayoría) con precisión. Eso parece ser solo el comienzo, con avances aún más impactantes en el audio y video de IA apuntando hacia un futuro en el que los mensajes de los líderes pueden ser traducidos en audio, con su propia voz.
La Traducción de Voz de Spotify para pódcast es una vista previa de cómo podría verse ese futuro, al menos para el audio. ¿En cuanto al video? No está muy atrasado.