# estrategIA’s first anniversary: testing ChatGPT’s new voice for political communication Author: Fernando Nieto Lobato Original publication: 2024-10-02 Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-53-un-ano-de-estrategia English URL: https://elcontemplador.github.io/estrategia-english/essays/053/ Status: Published translation This is a translation of the original Spanish essay published on 2 October 2024. Its claims, examples and forecasts retain that historical context. English publication: 2026-09-29 *Historical note: this article describes the features, restrictions and availability of ChatGPT’s advanced voice mode in October 2024.* A year ago, in October 2023 ([after an introductory issue published in the last week of September](https://open.substack.com/pub/estrategiabyaleph/p/bienvenidos-a-estrategia?r=2tyybn&utm_campaign=post&utm_medium=web)), [we launched estrategIA](https://open.substack.com/pub/estrategiabyaleph/p/estrategia-1?r=2tyybn&utm_campaign=post&utm_medium=web). **Today, on this first birthday, we want above all to thank you, dear readers, for your loyalty and support.** You make it possible for this venture—born above all from a desire to do something worthwhile and useful in helping us understand the great wave of artificial intelligence coming our way (which is barely wetting our feet for now, although we can see how enormous it is on the horizon)—to remain true to its purpose. It has grown considerably and now has well over 1,000 subscribers: at the time of writing, 759 on Substack and 317 on [LinkedIn Newsletters](https://www.linkedin.com/newsletters/estrategia-7201868200244834304/), where we began publishing four months ago. **Thank you all!** We hope you will stay with us for many more years, enjoying all the extraordinary innovations ahead. Let us remind you that, if you find estrategIA interesting, recommending it to colleagues and friends helps us enormously. **You can share it very easily through this [link](https://estrategiabyaleph.substack.com/).** **To celebrate the anniversary with a smile, [we would like to share a fun song we created in Suno](https://suno.com/song/0156cce7-1133-4fa1-8483-5749ebdb17ca), with lyrics by GPT-4o. Both were produced on the first attempt.** **Now to the substance.** The truth is that **our original idea for this birthday issue was to examine, as we have on other anniversaries, how artificial intelligence has evolved over the past year, particularly in relation to politics and government.** We would draw on material as interesting as [this survey](https://x.com/emollick/status/1838582600165670955) [from late August, shared by Ethan Mollick](https://x.com/emollick/status/1838582600165670955), which shows that, at least in the United States, people’s adoption of AI is beginning to take off at an unstoppable and very rapid pace. We will surely return in another issue to the interesting paper [“The Rapid Adoption of Generative AI”](https://static1.squarespace.com/static/60832ecef615231cedd30911/t/66f0c3fbabdc0a173e1e697e/1727054844024/BBD_GenAI_NBER_Sept2024.pdf). At the end of the year, we will review AI’s progress in 2024, [building on our review of the first six months](https://open.substack.com/pub/estrategiabyaleph/p/estrategia-39-balance-de-los-primeros?r=2tyybn&utm_campaign=post&utm_medium=web). [![Ethan Mollick’s September 2024 post on US generative AI adoption, with cropped research charts and a paper excerpt.](https://elcontemplador.github.io/estrategia-english/assets/images/a1043a65-d75b-4b79-b923-074c2076d970_900x934.png)](https://substackcdn.com/image/fetch/$s_!y3ui!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa1043a65-d75b-4b79-b923-074c2076d970_900x934.png) *Screenshot: Ethan Mollick’s post of 24 September 2024, with excerpts from the adoption study. The post is already in English; the charts and paper excerpts are partly cropped.* **Accessible transcript of the post:** > This August 2024 US survey of AI use confirms that AI is not just hype: > 1) AI adoption is crazy fast by historical standards > 2) It obviously useful to people in many industries. 25% people use GenAI at least 60 minutes a day at work, already > 3) AI is being used for many purposes The visible task chart asks, “In Which Specific Tasks Is AI Most Useful?” Its workplace results show the share ranking a task first or second: writing communications, 38.4%; performing administrative tasks, 26.8%; interpreting/translating/summarising, 23.0%; searching for facts or information, 17.7%; coding software, 13.8%; documentation or detailed instructions, 13.6%; generating/developing new ideas, 13.2%; support with customers/co-workers, 11.6%; data analysis/visualisation, 8.3%; tutoring or educational assistance, 3.5%. “Other” has no readable numerical label. The other panels compare adoption over time and across occupations and industries; their cropped labels and values are not reconstructed here. The linked paper supplies the full figures. That was the original idea, but **the release of [ChatGPT’s advanced voice model](https://www.technologyreview.com/2024/09/24/1104422/openai-released-its-advanced-voice-mode-to-more-people-heres-how-to-get-it/), [long awaited since its presentation in May](https://open.substack.com/pub/estrategiabyaleph/p/estrategia-33-google-vs-openai-una?r=2tyybn&utm_campaign=post&utm_medium=web), changed our plans.** It is not yet available in the EU, but we have been able to try it from Spain using a VPN. **This voice interface is revolutionary in its naturalness, despite arriving with its capabilities severely curtailed compared with the demonstration:** it still has no vision and does not integrate with GPTs; for safety reasons, it has been “forbidden” to sing, clone voices or recognise different people; and, for the time being, it has been deprived of memory. Even so, **it is undoubtedly the beginning of a shift towards a new way of communicating with machines, very close to how we speak with another person. As the example below will show, it also offers enormous possibilities for political communication and learning.** If you missed the presentation in May, here are [some recent examples](https://x.com/JStaatsCPA/status/1838984688917954621?t=3_DAA-4g4pHRRqtW7LGdFw&s=35) of the tool in general use. ## Testing GPT’s advanced voice mode to prepare for a debate When the model was presented more than four months ago, it occurred to me that one of the main uses of an AI that could speak like a human—and even interrupt or raise its voice, although in practice this is now much more restricted than the model itself allows—combined with GPT-4o’s reasoning capabilities, would be as **an excellent “sparring partner” for any politician preparing for a debate or a parliamentary speech involving a direct exchange with a colleague from another political camp.** **The audio we present below is our first take with this model.** Its imperfections are evident: for example, it takes on the role of a government member even though I have asked it to play the opposition. It does, however, correct itself very quickly. **We are presenting it as it is so that you can see how spontaneously the system works.** We had complex briefing notes from several parties on a range of subjects. But since advanced voice mode cannot yet be integrated with GPTs and has no dedicated long-term memory—two things that will probably change within a few months, allowing us to test the model much more thoroughly, as will the option of voice-specific “custom instructions”—**we preferred to create a simple set of arguments using [Perplexity](https://www.perplexity.ai/search/quiero-que-busques-en-internet-fJRZf5GNQai07uJW2E9.Yw).** We also chose **a subject** that, although very topical in Spain, is general and applicable in any country: **whether or not to reduce working hours.** **Below are the arguments we gave the AI (the opposition’s) and those we used ourselves (prepared for the government). They are followed by the exact, spontaneous, entirely unedited audio of our first mock “debate”.** Despite being a fairly spontaneous and very preliminary trial—this will improve enormously with prior customised instructions—it demonstrates many of the model’s possibilities: [![Perplexity’s Spanish briefing on arguments for and against reducing working hours in Spain; a full English translation follows.](https://elcontemplador.github.io/estrategia-english/assets/images/462fe851-e0ca-4d6d-9244-ebbb7a1b87eb_903x1600.png)](https://substackcdn.com/image/fetch/$s_!voEq!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F462fe851-e0ca-4d6d-9244-ebbb7a1b87eb_903x1600.png) ### Accessible translation of the historical Perplexity briefing *The following is the AI-generated text shown in the original screenshot, translated into English. It records the material used in this 2024 experiment. The numbered references are reproduced as displayed; their source list is not visible in the capture.* > Based on the available information, here is a summary of the arguments for and against reducing working hours in Spain: > > **Arguments in favour (trade unions and left-wing parties)** > > - **Better quality of life:** A shorter working day contributes to better mental and physical health for workers, reducing stress and exhaustion. [5, 6] > - **Increased productivity:** Several studies suggest that reducing working hours can increase productivity, as workers are more focused and efficient during their actual working time. [5, 6] > - **Lower unemployment:** Reducing working hours could help distribute employment more widely, allowing more staff to be hired to cover the reduced hours. [5, 7] > - **Work–life balance:** Employees would have more time for their families, leisure, training or rest, improving the balance between work and personal life. [1, 6] > - **Social justice:** Trade unions argue that this measure would help distribute the wealth generated by the country more equitably. [1, 2] > - **Positive impact on the economy and the environment:** Reducing working hours could have a positive effect on the country’s economy and the environment. [1, 2] > > **Arguments against (CEOE, employers’ organisations and right-wing parties)** > > - **Higher costs for businesses:** Businesses could face higher labour costs if they need to hire more staff to cover the reduced hours or pay overtime. [3, 5] > - **Loss of competitiveness:** In a globalised market, reducing working hours could affect Spanish businesses’ competitiveness relative to their international competitors. [3, 5] > - **Impact on specific sectors:** The adverse effects would be greater for smaller businesses and in sectors such as tourism, hospitality and retail, where face-to-face contact with customers is essential. [3] > - **Higher unit labour costs:** A compulsory reduction in working hours with wages held constant would increase unit labour costs, which could adversely affect businesses’ competitiveness and trends in employment and GDP. [3] > - **Competitive disadvantages:** Businesses operating in global markets may face competitive disadvantages if their labour costs rise relative to those in countries with longer working hours. [3] > - **Preference for collective bargaining:** CEOE and Cepyme reject a statutory reduction in working hours and point to collective bargaining as the framework for addressing a reduction in working time. [4] > > It is important to emphasise that these arguments represent each group’s general positions and that nuances and variations may exist within each perspective. ## Audio of the “debate training” with ChatGPT’s advanced voice mode *Archive note: the preserved local source announces an audio recording here but contains no playable audio or recording link. The [original Spanish newsletter](https://estrategiabyaleph.substack.com/p/estrategia-53-un-ano-de-estrategia) is retained as the source reference; the recording has not been reconstructed.* **Looking at this first attempt, we should remember what Sam Altman always says: “this is the stupidest AI you will ever use in your life”,** because these systems keep improving and improving. **That leads us to think that more capable models with this voice system, with fewer restrictions or in their next versions, may in the fairly near future become excellent tools for developing any politician’s public speaking and debating skills.** *Translation note: the quotation attributed to Sam Altman has been translated from the Spanish article; its original English wording has not been verified.* Fernando Nieto Lobato *Director of Digital Innovation at Institución Educativa ALEPH*