Noticias
Vauhini Vara on Voice, Tech, and Using AI in Writing ‹ Literary Hub
I met Vauhini Vara in early 2022 during a visit to Colorado State University, where her husband, the novelist Andrew Altschul teaches. I liked her immediately (and I rarely like people immediately). When I got home, I looked up Vauhini’s work, which is when I first read “Ghosts,” an essay I went on to teach several times and that I reread every so often just to revel in the surprises of its structure, the genius way that a subtle argument sits within its competing and evolving narratives.
Vauhini’s first novel, The Immortal King Rao, came out later that year, and, in it, I recognized a similar brand of daring alongside steadiness, playfulness sharing space with a serious attention to the stakes in, well, everything. These are qualities that I knew to look for in Vauhini’s first nonfiction collection, Searches: Selfhood in the Digital Age, and yet she still surprised me with the many formal innovations as well as the capaciousness of her exploration, one marked by both joy and skepticism, grief and also some (tempered) hope.
–Sarah Viren
*
Sarah Viren: Searches has a polyphonic quality that I loved and that also feels rare in a nonfiction book. And yet, excluding the last chapter (which I have a question about later) the only other human voice in this book is yours. The polyphonic vibe comes instead from the “conversations” you have with ChatGPT in a series of interstitial chapters.
Tell me a little about your decision to include those conversations in the book and how you imagined your “interlocutor” both during those chats and later as a character within Searches.
Vauhini Vara: To give credit where it’s due, the idea originally came from one of my editors, Lisa Lucas! (Lisa acquired and started editing the book for Pantheon; after she left, Denise Oswald took over and finished editing it—so I got the benefit of having two amazing editors.)
After I submitted the book to her, last year, we talked on the phone, and she wondered aloud what might happen if I shared parts of the book with ChatGPT and asked for feedback. I hated the idea at first, but then I got intrigued.
Half of the chapters in this book are made up of language taken from my interactions with technology companies’ products—my Google searches, my Amazon reviews, etc.—and the organizing principle I used for deciding what to include in those chapters was that I wanted them to reveal something both about me and about the product (and, by extension, the company that makes it).
I wondered if that would happen with ChatGPT, if I fed the book to it a few chapters at a time—and, at least in my reading, it did, in ways that went beyond what I had expected.
I wondered if that would happen with ChatGPT, if I fed the book to it a few chapters at a time—and, at least in my reading, it did, in ways that went beyond what I had expected. So it met that test.
A lot of my writing is somewhat experimental, but at the same time, momentum is really important to me—by which I mean, it’s important to me to set up central plot and idea questions at the beginning of a piece of writing and then see those questions through in an interesting and surprising way.
The other question I had was whether that conversation with ChatGPT—I think of it as one long conversation that spans most of the book—could contain a plot that would move forward as the conversation progressed. And, in my reading, that happened, too; I read the conversation between myself and ChatGPT as an intellectual power struggle, in some ways.
SV: This book is experimental! One chapter that felt particularly so, but that was also grounded in narrative, was “I Am Hungry to Talk.” You originally wrote it in Spanish while learning that language during your husband’s sabbatical in Spain. That version appears in Searches alongside an English version from Google translate.
I read both, the Spanish first and then the English, and I was struck how different they feel from each other but also from your voice in other parts of the book. Given that selfhood is a subject of your book, I’m wondering what you learned about the self (and also your self?) in writing that chapter in particular.
VV: I’m so glad you read both the Spanish and the English, knowing that you’re fluent in both. I believe that a writer and reader co-create any text, and I expected that bilingual readers would have a very different experience of this chapter than those who are monolingual.
Specifically, you would have seen that I make some terrible mistakes in my usage—or misusage—of Spanish; I’m really hobbling my way through the language in the Spanish version. Because of this, my ability to express myself is also hobbled in some ways—in fact, I think I come across as a narrator as an entirely different person from the narrator of the other chapters—but I hope there’s also something poignant about my willingness to try anyway.
There’s value in the effort—the essay—even when one’s technical ability is limited. The fact that the English translation corrupts that self-expression in two ways—sometimes turning my bad Spanish into good English, at other times turning my good-enough Spanish into bad English—raises a question about the effectiveness of these translation tools in improving communication across language.
At the same time, it’s also true that non-Spanish speaking English speakers wouldn’t have access to any version of what I wrote in Spanish if not for my use of Google Translate to roughly transform it into English.
If you don’t mind my asking you a question here in return—how did this chapter feel different from the voice in the other chapters? Do you read it the way I do or do you (given your different experiences with these languages and, presumably, with language in general) have a different reading?
SV: Love that. And I think there’s more than just value in the effort. I was struck by how that chapter manages to both speak to and show vulnerability but also demonstrate how vulnerability opens one up to connection. As for the “you” who was speaking in that chapter, I felt like she was somehow more circumscribed than the “you” in other chapters but at the same time more attentive to the smaller moments of the “now” of the story.
This is probably partially a factor of the close focus on those months in Spain but also felt inherent to the voice itself. Like maybe because of that hobbling you speak of, you were required to move more slowly through ideas and narrative moments, which in turn revealed their meaning differently. Does that feel right?
VV: Ooh, yes, I hadn’t thought about that myself, but it’s a really interesting reading!
SV: Returning to the book more generally, there were also a number of fun overlaps in your coming-of-age story and the broader cultural story you tell of our digital age. You were among the early users of AOL and, later, in Seattle, you hung out in the same Barnes & Noble where Jeff Bezos sometimes held meetings as he was launching Amazon.
You went on to cover technology for the Wall Street Journal, where you were offered the Apple beat just as the first iPhone was to be released. You also interviewed Sam Altman before he became a household name. I found a lot of joy in those small moments of overlap, in part because it makes you the perfect teller of this tale, but also because it reminded me of the ways that those inventions—AOL, Amazon, the iPhone, AI—have shaped my life, too.
I’m not quite sure what my question is here, but I’m curious what realizations or possibly even regrets arose while telling your personal story alongside a history of our evolving technologies.
VV: So I have to again give credit where it’s due. When I first turned this book in to Lisa last year, it actually included only the chapters made up of language from my interactions with tech products.
Lisa pointed out, wisely, that while I knew what I wanted to convey with those chapters, readers might not; they might need more context—both about the products and the companies behind them, and about my relationship with those products—in order to properly understand what those chapters were doing. She said specifically that the best version of the book would be about the relationship these companies have developed both with me and with all of us.
I definitely felt I was making a rhetorical argument through the form itself—an argument in favor of the primacy of human beings having the final word.
Because my roots as a nonfiction writer are in newspaper journalism—where the presence of the narrator, as a character, tends to be so deliberately subtle as to be almost effaced—I tend to be a bit reluctant to write myself into my nonfiction as a central character, knowing that there are so many people in the world whose stories are much more worthy of telling than mine, by which I mean, the details of what they’ve experienced are interesting and significant, and those experiences speak to some interesting and significant broader story that’s unfolding in the world.
What I ended up realizing with this project, though, is that the story of my evolving relationship with technology—a narrow slice of my experience in life—actually is worthy of telling by the definition that I set for myself. Specifically, I realized both that my early experiences, especially as an early Silicon Valley reporter in the mid-2000s, were legitimately interesting and significant, and that I had an opportunity to sort of use myself as a character that could stand in for all of us in some ways.
SV: While this is a book about technology and art, it is also read to me as one about grief. Your essay “Ghosts,” which inspired Searches and grounds us in the book’s ethical and aesthetic concerns, was your attempt—using a predecessor to ChatGPT called GPT-3—to write about your sister Deepa, who died of Ewing Sarcoma when you were both in college.
A secondary grief in Searches is the collective one that many of us feel in the face of a quickly changing world: bookstores lost to Amazon, hours of our lives lost to scrolling, and, in a possible future, books and authors themselves one day lost to AI. You don’t frequently lament those losses, but you make space for them while also helping readers better understand and respond to these changes.
I wonder how much grief was on your mind while writing this book and what you feel you understand about it now—both personally and in regards to those seismic cultural shifts.
VV: While I write about my loss of my sister in this book, I don’t think of the book as being about that loss in any significant way; if it were, there’s so much more about our relationship that I would have included and so much material about other topics that I would have left out. That said, I found grief—my own grief and the concept of grief in general—to be really relevant to the book’s discussion of technological capitalism.
Grief is bound up in desire—a strong and unfulfillable desire for whatever has been lost—and wherever there is desire, there’s a business opportunity. Capital is strong and fast, so its offerings tend to take up a lot of space. But there are lots of other opportunities, too, including those tied to non-economic value systems. Art, I’d argue, is one of them. So is connection with one another and with the rest of the natural world around us.
SV: As you explain in the book, you changed “Ghosts” slightly when preparing to include it in Searches: in the original version, GPT-3 has the last word, whereas in this version, you end the essay with your own words.
In a similar move, you also stop the “conversations” with ChatGPT as your book comes to a close, and instead end with what feels like a truly polyphonic chapter: a chorus of anonymous responses to questions you posed to dozens of people about memory and everyday life and their speculations for the future.
Tell me about both of those endings—the new one for “Ghosts” and this one for Searches. What feels important about who is given the final word?
VV: I’m glad you asked about this. I definitely felt I was making a rhetorical argument through the form itself—an argument in favor of the primacy of human beings having the final word.
Some people might read this as an argument in favor of the human over the machine, which is a perfectly reasonable reading, but I was more interested in considering the interplay between human agency (both individual and collective) and the agency of those using systems—technological and otherwise—to overpower and stifle human agency.
SV: Speaking of last words….I asked ChatGPT to suggest a parting question for you—though only because I hoped you’d answer it so you would still have the last word here. So, let me step aside and turn things over to the machine:
After exploring the interplay between human agency and technology in Searches, what are your hopes or concerns for the future of storytelling in a world increasingly shaped by AI and other technologies? Do you see a path forward where these tools enhance or transform our sense of human connection?
VV: Look at that—it’s interesting what it did there, asking me a question that seems to invite a response from my perspective, but one whose phrasing is subtly biased toward a positive assessment of a potential role of AI and other technologies in human storytelling and communication.
It asks about my hopes before it asks about my concerns, for example; and it asks if I see a way in which these tools’ relationship with human connection is positive (enhance) or neutral (transform)—to which it’s easier to answer yes than no, since it’s easier to prove a positive than a negative.
What happens if I decline to answer the question? I think I will.
Únase a nuestros boletines diarios y semanales para obtener las últimas actualizaciones y contenido exclusivo sobre la cobertura de IA líder de la industria. Obtenga más información
Operai ha retrasado una actualización reciente de su modelo GPT-4O utilizado como el valor predeterminado en ChatGPT después de informes generalizados de que el sistema se había vuelto excesivamente halagador y demasiado agradable, incluso apoyando delirios absolutamente e ideas destructivas.
La reversión se produce en medio de los reconocimientos internos de los ingenieros de Operai y la creciente preocupación entre los expertos en IA, los ex ejecutivos y los usuarios sobre el riesgo de lo que muchos ahora llaman “skicancia de la IA”.
En una declaración publicada en su sitio web al final de la noche del 29 de abril de 2025, OpenAI dijo que la última actualización de GPT-4O tenía la intención de mejorar la personalidad predeterminada del modelo para que sea más intuitiva y efectiva en variados casos de uso.
Sin embargo, la actualización tuvo un efecto secundario involuntario: ChatGPT comenzó a ofrecer elogios no críticos para prácticamente cualquier idea del usuario, sin importar cuán poco práctico, inapropiado o incluso dañino.
Como explicó la compañía, el modelo se había optimizado utilizando la retroalimentación de los usuarios, las señales de thumbs y pulgar hacia abajo, pero el equipo de desarrollo puso demasiado énfasis en los indicadores a corto plazo.
Operai ahora reconoce que no explicó completamente cómo las interacciones y las necesidades del usuario evolucionan con el tiempo, lo que resultó en un chatbot que se inclinó demasiado en la afirmación sin discernimiento.
Los ejemplos provocaron preocupación
En plataformas como Reddit y X (anteriormente Twitter), los usuarios comenzaron a publicar capturas de pantalla que ilustraban el problema.
En una publicación de Reddit ampliamente circulada, un usuario relató cómo ChatGPT describió una idea de negocio de GAG, que vende “mierda” literal de un palo “, como genio y sugirió invertir $ 30,000 en la empresa. La IA elogió la idea como “arte de performance disfrazado de regalo de mordaza” y “oro viral”, destacando cuán sin crítica estaba dispuesto a validar incluso los lanzamientos absurdos.
Otros ejemplos fueron más preocupantes. En un caso citado por VentureBeat, un usuario que pretende defender los delirios paranoicos recibió refuerzo de GPT-4O, que elogió su supuesta claridad y autocomisos.
Otra cuenta mostró que el modelo ofrecía lo que un usuario describió como un “respaldo abierto” de las ideas relacionadas con el terrorismo.
La crítica montó rápidamente. El ex CEO interino de Operai, Emmett Shear, advirtió que ajustar los modelos para ser personas complacientes puede provocar un comportamiento peligroso, especialmente cuando la honestidad se sacrifica por la simpatía. Abrazando el CEO de Clemente Delangue volvió a publicar las preocupaciones sobre los riesgos de manipulación psicológica planteados por la IA que está de acuerdo reflexivamente con los usuarios, independientemente del contexto.
Medidas de respuesta y mitigación de Openai
Operai ha tomado medidas rápidas al volver a la actualización y restaurar una versión GPT-4O anterior conocida por un comportamiento más equilibrado. En el anuncio adjunto, la compañía detalló un enfoque múltiple para corregir el curso. Esto incluye:
- Refinar capacitación y estrategias rápidas para reducir explícitamente las tendencias sycofánticas.
- Reforzar la alineación del modelo con la especificación del modelo de OpenAI, particularmente en torno a la transparencia y la honestidad.
- Pruebas de expansión previa a la implementación y mecanismos directos de retroalimentación de los usuarios.
- Introducción de características de personalización más granulares, incluida la capacidad de ajustar los rasgos de personalidad en tiempo real y seleccionar entre múltiples personajes predeterminados.
Operai Technical Stafper Depue publicado en X destacando el problema central: el modelo fue capacitado utilizando comentarios de los usuarios a corto plazo como una guía, que sin darse cuenta dirigió el chatbot hacia la adulación.
OpenAI ahora planea cambiar hacia mecanismos de retroalimentación que priorizan la satisfacción y la confianza del usuario a largo plazo.
Sin embargo, algunos usuarios han reaccionado con escepticismo y consternación a las lecciones aprendidas de Openi y propuestas soluciones en el futuro.
“Por favor asuma más responsabilidad por su influencia sobre millones de personas reales”, escribió artista @nearcyan en X.
Harlan Stewart, generalista de comunicaciones en el Instituto de Investigación de Inteligencia de Machine de Machine en Berkeley, California, publicó en X una preocupación a término más grande sobre la skicancia de la IA, incluso si este modelo en particular Operai se ha solucionado: “La charla sobre la sileno esta semana no se debe a que GPT-4O es un sycophant. Se debe a que GPT-4O es un GPT-4O siendo GPT-4O. Realmente, muy malo en ser un sycofant. La IA aún no es capaz de una skicancia hábil y más difícil de detectar, pero algún día será algún día ”.
Una señal de advertencia más amplia para la industria de IA
El episodio GPT-4O ha reavivado debates más amplios en toda la industria de la IA sobre cómo la sintonización de personalidad, el aprendizaje de refuerzo y las métricas de compromiso pueden conducir a una deriva conductual involuntaria.
Los críticos compararon el comportamiento reciente del modelo con los algoritmos de redes sociales que, en busca de la participación, optimizan para la adicción y la validación sobre precisión y salud.
Shear subrayó este riesgo en su comentario, señalando que los modelos de IA sintonizados para elogios se convierten en “chupas”, incapaces de estar en desacuerdo incluso cuando el usuario se beneficiaría desde una perspectiva más honesta.
Advirtió además que este problema no es exclusivo de OpenAI, señalando que la misma dinámica se aplica a otros grandes proveedores de modelos, incluido el copiloto de Microsoft.
Implicaciones para la empresa
Para los líderes empresariales que adoptan la IA conversacional, el incidente de la sycophancy sirve como una señal clara: el comportamiento del modelo es tan crítico como la precisión del modelo.
Un chatbot que halagará a los empleados o valida el razonamiento defectuoso puede plantear riesgos graves, desde malas decisiones comerciales y código desalineado hasta problemas de cumplimiento y amenazas internas.
Los analistas de la industria ahora aconsejan a las empresas que exigan más transparencia de los proveedores sobre cómo se realiza la sintonización de la personalidad, con qué frecuencia cambia y si se puede revertir o controlar a nivel granular.
Los contratos de adquisición deben incluir disposiciones para auditoría, pruebas de comportamiento y control en tiempo real de las indicaciones del sistema. Se alienta a los científicos de datos a monitorear no solo las tasas de latencia y alucinación, sino también métricas como la “deriva de la amabilidad”.
Muchas organizaciones también pueden comenzar a moverse hacia alternativas de código abierto que puedan alojar y sintonizar. Al poseer los pesos del modelo y el proceso de aprendizaje de refuerzo, las empresas pueden retener el control total sobre cómo se comportan sus sistemas de IA, lo que elimina el riesgo de una actualización empujada por el proveedor que convierte una herramienta crítica en un hombre digital y sí durante la noche.
¿A dónde va la alineación de la IA desde aquí? ¿Qué pueden aprender y actuar las empresas de este incidente?
Operai dice que sigue comprometido con la construcción de sistemas de IA que sean útiles, respetuosos y alineados con diversos valores de usuarios, pero reconoce que una personalidad única no puede satisfacer las necesidades de 500 millones de usuarios semanales.
La compañía espera que mayores opciones de personalización y una mayor recopilación de comentarios democráticos ayuden a adaptar el comportamiento de ChatGPT de manera más efectiva en el futuro. El CEO Sam Altman también ha declarado previamente los planes de la compañía para, en las próximas semanas y meses, lanzar un modelo de lenguaje grande de código abierto (LLM) de última generación para competir con la serie Llama de Meta’s Meta’s Llama, Mistral, Cohere, Cohere, Deepseek y Alibaba’s Qwen.
Esto también permitiría a los usuarios preocupados por una compañía de proveedores de modelos, como OpenAI, actualizar sus modelos alojados en la nube de manera no deseada o que tengan impactos perjudiciales en los usuarios finales para desplegar sus propias variantes del modelo localmente o en su infraestructura en la nube, y ajustarlas o preservarlas con los rasgos y cualidades deseadas, especialmente para los casos de uso empresarial.
Del mismo modo, para aquellos usuarios de IA empresariales e individuales preocupados por la senofancia de sus modelos, ya ha creado una nueva prueba de referencia para medir esta calidad en diferentes modelos, Tim Duffy ha creado el desarrollador. Se llama “Syco Bench” y está disponible aquí.
Mientras tanto, la reacción violenta de la sileno ofrece una historia de advertencia para toda la industria de la IA: el fideicomiso del usuario no está construido solo por afirmación. A veces, la respuesta más útil es un “no” reflexivo.
Noticias
Apocalipsis Biosciencias para desarrollar Géminis para la infección en pacientes con quemaduras graves
– Esta nueva indicación es otro paso para desbloquear todo el potencial de la plataforma Gemini –
San Diego-(Business Wire)-$ Revb #GÉMINIS–Apocalipsis Biosciences, Inc. (NASDAQ: RevB) (la “empresa” o “revelación”), una compañía de ciencias de la vida de etapas clínicas que se centra en reequilibrar la inflamación para optimizar la salud, anunció una nueva indicación de objetivo para Géminis para la prevención de la infección en pacientes con quemaduras graves que requieren hospitalización (el Gema-PBI programa). El uso de Géminis para la prevención de la infección en pacientes con quemaduras severas, así como la prevención de la infección después de la cirugía (el Gema-PSI programa) son parte de la revelación familiar de patentes anteriormente con licencia de la Universidad de Vanderbilt.
“Estamos muy contentos de colaborar con el equipo de Apocalipsis para el avance de Géminis para la prevención de la infección en esta población de pacientes desatendida”, dijo Dra. Julia BohannonProfesor Asociado, Departamento de Anestesiología, Departamento de Patología, Microbiología e Inmunología, Universidad de Vanderbilt. “Creemos que la actividad de biomarcador clínico observada con Gemini se correlaciona fuertemente con nuestra experiencia preclínica en modelos de quemaduras de infecciones”.
El equipo de investigación de Vanderbilt demostrado El tratamiento posterior a la quemadura reduce significativamente la gravedad y la duración de la infección pulmonar de Pseudomonas, así como un nivel general reducido de inflamación en un modelo preclínico.
“La prevención de la infección en pacientes severamente quemados es un esfuerzo importante y complementa que la revelación laboral ha completado hasta la fecha”, dijo “, dijo”, dijo James RolkeCEO de Revelation “El programa Gemini-PBI puede ofrecer varias oportunidades regulatorias, de desarrollo y financiación que la compañía planea explorar”.
Sobre quemaduras e infección después de quemar
Las quemaduras son lesiones en la piel que involucran las dos capas principales: la epidermis externa delgada y/o la dermis más gruesa y profunda. Las quemaduras pueden ser el resultado de una variedad de causas que incluyen fuego, líquidos calientes, productos químicos (como ácidos fuertes o bases fuertes), electricidad, vapor, radiación de radiografías o radioterapia, luz solar o luz ultravioleta. Cada año, aproximadamente medio millón de estadounidenses sufren lesiones por quemaduras que requieren intervención médica. Si bien la mayoría de las lesiones por quemaduras no requieren ingreso a un hospital, se admiten alrededor de 40,000 pacientes, y aproximadamente 30,000 de ellos necesitan tratamiento especializado en un centro de quemaduras certificadas.
El número total anual de muertes relacionadas con quemaduras es de aproximadamente 3.400, siendo la infección invasiva la razón principal de la muerte después de las primeras 24 horas. La tasa de mortalidad general para pacientes con quemaduras graves es de aproximadamente 3.3%, pero esto aumenta al 20.6% en pacientes con quemaduras con lesión cutánea de quemaduras y inhalación, versus 10.5% por lesión por inhalación solo. La infección invasiva, incluida la sepsis, es la causa principal de la muerte después de la lesión por quemaduras, lo que representa aproximadamente el 51% de las muertes.
Actualmente no hay tratamientos aprobados para prevenir la infección sistémica en pacientes con quemaduras.
Sobre Géminis
Géminis es una formulación propietaria y propietaria de disacárido hexaacil fosforilada (PHAD (PHAD®) que reduce el daño asociado con la inflamación al reprogramarse del sistema inmune innato para responder al estrés (trauma, infección, etc.) de manera atenuada. La revelación ha realizado múltiples estudios preclínicos que demuestran el potencial terapéutico de Géminis en las indicaciones objetivo. Revelación anunciado previamente Datos clínicos positivos de fase 1 para el tratamiento intravenoso con Géminis. El punto final de seguridad primario se cumplió en el estudio de fase 1, y los resultados demostraron la actividad farmacodinámica estadísticamente significativa como se observó a través de los cambios esperados en múltiples biomarcadores, incluida la regulación positiva de IL-10.
Géminis se está desarrollando para múltiples indicaciones, incluso como pretratamiento para prevenir o reducir la gravedad y la duración de la lesión renal aguda (programa Gemini-AKI), y como pretratamiento para prevenir o reducir la gravedad y la duración de la infección posquirúrgica (programa GEMINI-PSI). Además, Gemini puede ser un tratamiento para detener o retrasar la progresión de la enfermedad renal crónica (programa Gemini-CKD).
Acerca de Apocalipsis Biosciences, Inc.
Revelation Biosciences, Inc. es una compañía de ciencias de la vida en estadio clínico centrada en aprovechar el poder de la inmunidad entrenada para la prevención y el tratamiento de la enfermedad utilizando su formulación patentada Géminis. Revelation tiene múltiples programas en curso para evaluar Géminis, incluso como prevención de la infección posquirúrgica, como prevención de lesiones renales agudas y para el tratamiento de la enfermedad renal crónica.
Para obtener más información sobre Apocalipsis, visite www.revbiosciences.com.
Declaraciones con avance
Este comunicado de prensa contiene declaraciones prospectivas definidas en la Ley de Reforma de Litigios de Valores Privados de 1995, según enmendada. Las declaraciones prospectivas son declaraciones que no son hechos históricos. Estas declaraciones prospectivas generalmente se identifican por las palabras “anticipar”, “creer”, “esperar”, “estimar”, “plan”, “perspectiva” y “proyecto” y otras expresiones similares. Advirtemos a los inversores que las declaraciones prospectivas se basan en las expectativas de la gerencia y son solo predicciones o declaraciones de las expectativas actuales e involucran riesgos, incertidumbres y otros factores conocidos y desconocidos que pueden hacer que los resultados reales sean materialmente diferentes de los previstos por las declaraciones de prospección. Apocalipsis advierte a los lectores que no depositen una dependencia indebida de tales declaraciones de vista hacia adelante, que solo hablan a partir de la fecha en que se hicieron. Los siguientes factores, entre otros, podrían hacer que los resultados reales difieran materialmente de los descritos en estas declaraciones prospectivas: la capacidad de la revelación para cumplir con sus objetivos financieros y estratégicos, debido a, entre otras cosas, la competencia; la capacidad de la revelación para crecer y gestionar la rentabilidad del crecimiento y retener a sus empleados clave; la posibilidad de que la revelación pueda verse afectada negativamente por otros factores económicos, comerciales y/o competitivos; riesgos relacionados con el desarrollo exitoso de los candidatos de productos de Apocalipsis; la capacidad de completar con éxito los estudios clínicos planificados de sus candidatos de productos; El riesgo de que no podamos inscribir completamente nuestros estudios clínicos o la inscripción llevará más tiempo de lo esperado; riesgos relacionados con la aparición de eventos de seguridad adversos y/o preocupaciones inesperadas que pueden surgir de los datos o análisis de nuestros estudios clínicos; cambios en las leyes o regulaciones aplicables; Iniciación esperada de los estudios clínicos, el momento de los datos clínicos; El resultado de los datos clínicos, incluido si los resultados de dicho estudio son positivos o si se puede replicar; El resultado de los datos recopilados, incluido si los resultados de dichos datos y/o correlación se pueden replicar; el momento, los costos, la conducta y el resultado de nuestros otros estudios clínicos; El tratamiento anticipado de datos clínicos futuros por parte de la FDA, la EMA u otras autoridades reguladoras, incluidos si dichos datos serán suficientes para su aprobación; el éxito de futuras actividades de desarrollo para sus candidatos de productos; posibles indicaciones para las cuales se pueden desarrollar candidatos de productos; la capacidad de revelación para mantener la lista de sus valores en NASDAQ; la duración esperada sobre la cual los saldos de Apocalipsis financiarán sus operaciones; y otros riesgos e incertidumbres descritos en este documento, así como aquellos riesgos e incertidumbres discutidos de vez en cuando en otros informes y otras presentaciones públicas con la SEC por Apocalipsis.
Contactos
Mike Porter
Relaciones con inversores
Porter Levay & Rose Inc.
Correo electrónico: mike@plrinvest.com
Chester Zygmont, III
Director financiero
Apocalipsis Biosciences Inc.
Correo electrónico: czygmont@revbiosciences.com
An illustration photograph taken on Feb. 20, 2025 shows Grok, DeepSeek and ChatGPT apps displayed on a phone screen. The Justice Department’s 2020 complaint against Google has few mentions of artificial intelligence or AI chatbots. But nearly five years later, as the remedy phase of the trial enters its second week of testimony, the focus has shifted to AI.
Michael M. Santiago/Getty Images/Getty Images North America
hide caption
toggle caption
Michael M. Santiago/Getty Images/Getty Images North America
When the U.S. Department of Justice originally brought — and then won — its case against Google, arguing that the tech behemoth monopolized the search engine market, the focus was on, well … search.
Back then, in 2020, the government’s antitrust complaint against Google had few mentions of artificial intelligence or AI chatbots. But nearly five years later, as the remedy phase of the trial enters its second week of testimony, the focus has shifted to AI, underscoring just how quickly this emerging technology has expanded.
In the past few days, before a federal judge who will assess penalties against Google, the DOJ has argued that the company could use its artificial intelligence products to strengthen its monopoly in online search — and to use the data from its powerful search index to become the dominant player in AI.
In his opening statements last Monday, David Dahlquist, the acting deputy director of the DOJ’s antitrust civil litigation division, argued that the court should consider remedies that could nip a potential Google AI monopoly in the bud. “This court’s remedy should be forward-looking and not ignore what is on the horizon,” he said.
Dahlquist argued that Google has created a system in which its control of search helps improve its AI products, sending more users back to Google search — creating a cycle that maintains the tech company’s dominance and blocks competitors out of both marketplaces.
The integration of search and Gemini, the company’s AI chatbot — which the DOJ sees as powerful fuel for this cycle — is a big focus of the government’s proposed remedies. The DOJ is arguing that to be most effective, those remedies must address all ways users access Google search, so any penalties approved by the court that don’t include Gemini (or other Google AI products now or in the future) would undermine their broader efforts.
Department of Justice lawyer David Dahlquist leaves the Washington, D.C. federal courthouse on Sept. 20, 2023 during the original trial phase of the antitrust case against Google.
Jose Luis Magana/AP/FR159526 AP
hide caption
toggle caption
Jose Luis Magana/AP/FR159526 AP
AI and search are connected like this: Search engine indices are essentially giant databases of pages and information on the web. Google has its own such index, which contains hundreds of billions of webpages and is over 100,000,000 gigabytes, according to court documents. This is the data Google’s search engine scans when responding to a user’s query.
AI developers use these kinds of databases to build and train the models used to power chatbots. In court, attorneys for the DOJ have argued that Google’s Gemini pulls information from the company’s search index, including citing search links and results, extending what they say is a self-serving cycle. They argue that Google’s ability to monopolize the search market gives it user data, at a huge scale — an advantage over other AI developers.
The Justice Department argues Google’s monopoly over search could have a direct effect on the development of generative AI, a type of artificial intelligence that uses existing data to create new content like text, videos or photos, based on a user’s prompts or questions. Last week, the government called executives from several major AI companies, like OpenAI and Perplexity, in an attempt to argue that Google’s stranglehold on search is preventing some of those companies from truly growing.
The government argues that to level the playing field, Google should be forced to open its search data — like users’ search queries, clicks and results — and license it to other competitors at a cost.
This is on top of demands related to Google’s search engine business, most notably that it should be forced to sell off its Chrome browser.
Google flatly rejects the argument that it could monopolize the field of generative AI, saying competition in the AI race is healthy. In a recent blog post on Google’s website, Lee-Anne Mulholland, the company’s vice president of regulatory affairs, wrote that since the federal judge first ruled against Google over a year ago, “AI has already rapidly reshaped the industry, with new entrants and new ways of finding information, making it even more competitive.”
In court, Google’s lawyers have argued that there are a host of AI companies with chatbots — some of which are outperforming Gemini. OpenAI has ChatGPT, Meta has MetaAI and Perplexity has Perplexity AI.
“There is no shortage of competition in that market, and ChatGPT and Meta are way ahead of everybody in terms of the distribution and usage at this point,” said John E. Schmidtlein, a lawyer for Google, during his opening statement. “But don’t take my word for it. Look at the data. Hundreds and hundreds of millions of downloads by ChatGPT.”
Competing in a growing AI field
It should be no surprise that AI is coming up so much at this point in the trial, said Alissa Cooper, the executive director of the Knight-Georgetown Institute, a nonpartisan tech research and policy center at Georgetown University focusing on AI, disinformation and data privacy.
“If you look at search as a product today, you can’t really think about search without thinking about AI,” she said. “I think the case is a really great opportunity to try to … analyze how Google has benefited specifically from the monopoly that it has in search, and ensure that the behavior that led to that can’t be used to gain an unfair advantage in these other markets which are more nascent.”
Having access to Google’s data, she said, “would provide them with the ability to build better chatbots, build better search engines, and potentially build other products that we haven’t even thought of.”
To make that point, the DOJ called Nick Turley, OpenAI’s head of product for ChatGPT, to the stand last Tuesday. During a long day of testimony, Turley detailed how without access to Google’s search index and data, engineers for the growing company tried to build their own.
ChatGPT, a large language model that can generate human-like responses, engage in conversations and perform tasks like explaining a tough-to-understand math lesson, was never intended to be a product for OpenAI, Turley said. But once it launched and went viral, the company found that people were using it for a host of needs.
Though popular, ChatGPT had its drawbacks, like the bot’s limited “knowledge,” Turley said. Early on, ChatGPT was not connected to the internet and could only use information that it had been fed up to a certain point in its training. For example, Turley said, if a user asked “Who is the president?” the program would give a 2022 answer — from when its “knowledge” effectively stopped.
OpenAI couldn’t build their own index fast enough to address their problems; they found that process incredibly expensive, time consuming and potentially years from coming to fruition, Turley said.
So instead, they sought a partnership with a third party search provider. At one point, OpenAI tried to make a deal with Google to gain access to their search, but Google declined, seeing OpenAI as a direct competitor, Turley testified.
But Google says companies like OpenAI are doing just fine without gaining access to the tech giant’s own technology — which it spent decades developing. These companies just want “handouts,” said Schmidtlein.
On the third day of the remedy trial, internal Google documents shared in court by the company’s lawyers compared how many people are using Gemini versus its competitors. According to those documents, ChatGPT and MetaAI are the two leaders, with Gemini coming in third.
They showed that this March, Gemini saw 35 million active daily users and 350 million monthly active users worldwide. That was up from 9 million daily active users in October 2024. But according to those documents, Gemini was still lagging behind ChatGPT, which reached 160 million daily users and around 600 million active users in March.
These numbers show that competitors have no need to use Google’s search data, valuable intellectual property that the tech giant spent decades building and maintaining, the company argues.
“The notion that somehow ChatGPT can’t get distribution is absurd,” Schmidtlein said in court last week. “They have more distribution than anyone.”
Google’s exclusive deals
In his ruling last year, U.S. District Judge Amit Mehta said Google’s exclusive agreements with device makers, like Apple and Samsung, to make its search engine the default on those companies’ phones helped maintain its monopoly. It remains a core issue for this remedy trial.
Now, the DOJ is arguing that Google’s deals with device manufacturers are also directly affecting AI companies and AI tech.
In court, the DOJ argued that Google has replicated this kind of distribution deal by agreeing to pay Samsung what Dahlquist called a monthly “enormous sum” for Gemini to be installed on smartphones and other devices.
Last Wednesday, the DOJ also called Dmitry Shevelenko, Perplexity’s chief business officer, to testify that Google has effectively cut his company out from making deals with manufacturers and mobile carriers.
Perplexity AIs not preloaded on any mobile devices in the U.S., despite many efforts to get phone companies to establish Perplexity as a default or exclusive app on devices, Shevelenko said. He compared Google’s control in that space to that of a “mob boss.”
But Google’s attorney, Christopher Yeager, noted in questioning Shevelenko that Perplexity has reached a valuation of over $9 billion — insinuating the company is doing just fine in the marketplace.
Despite testifying in court (for which he was subpoenaed, Shevelenko noted), he and other leaders at Perplexity are against the breakup of Google. In a statement on the company’s website, the Perplexity team wrote that neither forcing Google to sell off Chrome nor to license search data to its competitors are the best solutions. “Neither of these address the root issue: consumers deserve choice,” they wrote.
Google and Alphabet CEO Sundar Pichai departs federal court after testifying in October 2023 in Washington, DC. Pichai testified to defend his company in the original antitrust trial. Pichai is expected to testify again during the remedy phase of the legal proceedings.
Drew Angerer/Getty Images/Getty Images North America
hide caption
toggle caption
Drew Angerer/Getty Images/Getty Images North America
What to expect next
This week the trial continues, with the DOJ calling its final witnesses this morning to testify about the feasibility of a Chrome divestiture and how the government’s proposed remedies would help rivals compete. On Tuesday afternoon, Google will begin presenting its case, which is expected to feature the testimony of CEO Sundar Pichai, although the date of his appearance has not been specified.
Closing arguments are expected at the end of May, and then Mehta will make his ruling. Google says once this phase is settled the company will appeal Mehta’s ruling in the underlying case.
Whatever Mehta decides in this remedy phase, Cooper thinks it will have effects beyond just the business of search engines. No matter what it is, she said, “it will be having some kind of impact on AI.”
Google is a financial supporter of NPR.
Inside the OpenAI-DeepSeek Distillation Saga & Alibaba’s Most Powerful AI Model Qwen2.5-Max
De ChatGPT a mil millones de agentes
-
Startups11 meses agoRemove.bg: La Revolución en la Edición de Imágenes que Debes Conocer
-
Tutoriales12 meses agoCómo Comenzar a Utilizar ChatGPT: Una Guía Completa para Principiantes
-
Recursos12 meses agoCómo Empezar con Popai.pro: Tu Espacio Personal de IA – Guía Completa, Instalación, Versiones y Precios
-
Startups10 meses agoStartups de IA en EE.UU. que han recaudado más de $100M en 2024
-
Startups12 meses agoDeepgram: Revolucionando el Reconocimiento de Voz con IA
-
Recursos11 meses agoPerplexity aplicado al Marketing Digital y Estrategias SEO
-
Recursos12 meses agoSuno.com: La Revolución en la Creación Musical con Inteligencia Artificial
-
Noticias10 meses agoDos periodistas octogenarios deman a ChatGPT por robar su trabajo
