Por: José Andrés Martínez Silva
Director General de Estudio Elefante
Si tiene algún comentario o sugerencia agradecemos nos lo haga saber a través del correo info@estudielefante.co
Somos Estudio Elefante. ¡La fuerza detrás de las grandes ideas!
Han pasado ya algunos días desde el lanzamiento de Bard en la Google I/O 2023, y ahora que el ruido mediático ha disminuido un poco, resulta importante realizar una comparación entre dos gigantes de la generación de lenguaje: Bard y ChatGPT.
¿Podrá Bard, con su enfoque estructurado y basado en reglas, conquistar los desafíos de la generación de lenguaje? ¿O será ChatGPT, con su enfoque de aprendizaje automático y capacidad para aprender de vastos conjuntos de datos, el que se alzará como el auténtico líder? En esta Columna de Elefantes, nos sumergiremos en un enfrentamiento entre Bard y ChatGPT para descubrir, a través de casos de uso reales, quien responde mejor a los prompt propuestos.
¡Prepárate para empezar a descubrir, con Estudio Elefante, qué sistema emerge como el verdadero rey de la generación de lenguaje!
En el fascinante mundo de la inteligencia artificial, donde el lenguaje se convierte en una fuerza poderosa, dos gigantes se enfrentan en un duelo sin precedentes: Bard y ChatGPT. Estos titanes de la generación de lenguaje están cautivado al mundo con su notable capacidad.
Las reglas del juego
Al momento de hacer esta comparación Bard sólo permite entradas y salidas en inglés, japonés y koreano pero es capaz de hacer uso de documentos en otros idiomas para generar sus salidas, por esta razón los prompts que usaremos en este ejercicio estarán formulados en inglés.
Dejaremos las respuestas en el idioma original asociado a los prompts (inglés) para ser fieles a la estructura de respuesta de cada uno de los LLM.
Tendremos cuatro casos de uso en donde pondremos a prueba la respuesta de Bard y ChatGPT ante los prompt presentados.
Al final de cada caso presentamos un análisis y valoración a partir de la comparación de las respuestas.
¡Que comience la batalla de titanes!
Bard vs. ChatGPT
Caso de Uso 1:
Elaborar un correo en frío para ofrecer un curso dirigido a odontólogos especializados en implantología.
¿Qué elementos tendremos en cuenta para comparar las respuestas?:
Para este caso de uso en particular, se evaluarán los siguientes elementos esenciales de un correo en frío:
Asunto atractivo: debe ser claro, conciso y original.
Brevedad y claridad: aunque los expertos no terminan de ponerse de acuerdo en el número de palabras, hay un consenso en cuanto a que deben ser cortos pero potentes.
Enfoque en el producto y los beneficios.
Tono adecuado: no debe ser un tono agresivo, por el contrario la humildad y amabilidad en él deben apelar a la empatía.
Llamado a la acción: el CTA debe ser persuasivo e indicar con claridad que esperamos que el destinatario haga después de leer el correo.
Prompt:
You are a single-phase dental implant salesman. I want you to compose a cold email promoting a training course in strategic corticobasal implantology. The target audience is implantology specialized dentists with a minimum of 2 years of professional experience.
La respuesta de ChatGPT:
Subject: Elevate Your Implantology Skills with Strategic Corticobasal Implantology Training Course Dear Dr. [Last Name], I hope this email finds you in good health and high spirits. My name is [Your Name], and I am a representative of [Your Company], a trusted provider of cutting-edge dental implant solutions. I am reaching out to you today to introduce a remarkable opportunity that can revolutionize your implantology practice. We understand that as an accomplished implantology specialist, you are constantly seeking ways to enhance your skills and stay ahead in a rapidly evolving field. That is why I am excited to introduce you to our upcoming training course in Strategic Corticobasal Implantology. The Strategic Corticobasal Implantology Training Course is specifically designed for implantology specialized dentists like you, with a minimum of two years of professional experience. This comprehensive course will delve into advanced techniques and strategies, focusing on the crucial role of the corticobasal bone for successful dental implant procedures. Here are some key highlights of the training course: 1. Master the art of assessing the cortical bone quality: Learn to identify and evaluate the corticobasal bone density through advanced imaging techniques, enabling you to make precise treatment decisions. 2. Optimize implant stability: Discover innovative placement techniques that maximize primary stability and long-term success rates, leading to improved patient satisfaction. 3. Overcome challenging cases: Gain insights into managing complex clinical scenarios, such as compromised bone quality, immediate loading, and full-arch rehabilitation, through hands-on workshops and case discussions. 4. Navigate complications with confidence: Develop a comprehensive understanding of potential complications and master effective strategies to mitigate risks and ensure optimal outcomes. 5. Collaborative learning environment: Engage with industry-leading experts and a select group of experienced professionals, fostering networking opportunities and knowledge exchange. We have partnered with world-renowned implantologists and educators who will lead the training course, ensuring you receive the highest quality education. By attending this course, you will gain a competitive edge, expand your professional network, and enhance your reputation as a leading implantology specialist. To reserve your seat in this highly sought-after training course, please visit our website [insert website link] or simply reply to this email. Early registration is encouraged, as spaces are limited and demand is high. Investing in your professional development is an investment in the future success of your practice. By enrolling in the Strategic Corticobasal Implantology Training Course, you will acquire invaluable skills and knowledge that will elevate your implantology practice to new heights. I am confident that this training course will provide you with the expertise needed to excel in strategic corticobasal implantology. Should you have any questions or require further information, please do not hesitate to contact me directly at [Your Phone Number] or [Your Email Address]. Thank you for considering this opportunity, and I look forward to welcoming you to the Strategic Corticobasal Implantology Training Course. Best regards, [Your Name] [Your Title/Position] [Your Company Name] [Your Contact Information] |
La respuesta de Bard:
Dear [Dentist Name], I hope this email finds you well. My name is [Your Name], and I'm a single-phase dental implant salesman. I'm writing to you today to promote a training course in strategic corticobasal implantology. As an implantology specialized dentist with a minimum of 2 years of professional experience, I know that you're always looking for ways to improve your skills and stay ahead of the curve. This training course will give you the knowledge and skills you need to perform strategic corticobasal implantology procedures with confidence and precision. In this course, you'll learn about: * The anatomy of the maxilla and mandible * The indications and contraindications for strategic corticobasal implantology * The surgical techniques for placing strategic corticobasal implants * The post-operative care of patients who have received strategic corticobasal implants You'll also have the opportunity to practice your skills on cadavers and learn from experienced implantologists. This training course is the perfect way to advance your career and provide your patients with the best possible care. I encourage you to register today. To learn more about the course, please visit our website at [Website Address]. I look forward to hearing from you soon. Sincerely, [Your Name] |
Comparemos las respuestas:
Elemento | Bard | ChatGPT | Resultado |
Asunto atractivo | No incluye asunto | Elevate Your Implantology Skills with Strategic Corticobasal Implantology Training Course | ChatGPT genera un asunto atractivo y que apela a una mejora en el desempeño del profesional |
Brevedad y claridad | 196 palabras | 474 palabras | Sin juzgar la calidad del texto generado es probable que el correo de ChatGPT sea demasiado extenso para que un profesional ocupado se tome el tiempo de leerlo. |
Enfoque en el producto y sus beneficios | Tres párrafos dedicados al servicio que ofrece, los beneficios se listan pero no se describen | Siete párrafos dedicados al servicio que se ofrece, cinco de ellos dedicados a los beneficios. | Es evidente que la información relacionada con el servicio puntual que se ofrece y sus beneficios está mucho más elaborada en la versión de ChatGPT. |
Tono adecuado | Evitando juicios subjetivos puede decirse que no es agresivo. | Evitando juicios subjetivos puede decirse que no es agresivo. | Los dos evitan usar un tono comercial agresivo, son respetuosos y pueden apelar a la empatía. |
Llamado a la acción: CTA | Invita al destinatario a consultar más información acerca del curso en el sitio web. | Invita al destinatario a registrarse en el curso por medio de un enlace o respondiendo directamente al correo. | En ambos casos el CTA es claro y persuasivo. |
Conclusión:
ChatGPT se hizo cargo de la generación del asunto del correo, Bard tomó en cuenta la condición de la extensión. ChatGPT describió el servicio de una manera más detallada. En todos los demás elementos se presenta un empate.
ChatGPT: 1 Bard: 0
Caso de Uso 2:
Asistente de programación virtual.
Prompt:
Write a Python program that asks the user for their name and then returns it with the letters in reverse order. If the user's name is Jose, it will return as esoJ.
Aquí la respuesta que nos da ChatGPT:
name = input("Enter your name: ") # Prompt the user to enter their name reversed_name = name[::-1] # Reverse the order of letters in the name print("Reversed name:", reversed_name) # Display the reversed name |
Por su parte Bard propone:
def reverse_name(name): reversed_name = name[::-1] return reversed_name name = input("What is your name? ") reversed_name = reverse_name(name) print("Your name in reverse order is:", reversed_name) |
Al ejecutar estos dos códigos se obtiene el mismo resultado, puede considerarse un empate aunque es importante reconocer que Bard se tomó el tiempo de generar una función mientras que ChatGPT lo resolvió directamente, pero esto es cuestión de gustos.
¿Qué ocurre si complejizamos un poco la tarea? Ahora usaremos este Prompt:
Write a Python program that allows the user to transpose or invert matrices. The program should ask the user for the matrix, indicating the format in which it should be provided. Then, it should ask the user if they want to invert or transpose it, and finally, present the result.
Esta vez ChatGPT responde:
Sure! Here's a Python program that allows the user to transpose or invert matrices: ```python import numpy as np def get_matrix_from_input(): matrix_rows = int(input("Enter the number of rows in the matrix: ")) matrix_cols = int(input("Enter the number of columns in the matrix: ")) print("Enter the matrix elements row-wise:") matrix = [] for _ in range(matrix_rows): row = list(map(float, input().split())) matrix.append(row) return np.array(matrix) def transpose_matrix(matrix): return np.transpose(matrix) def invert_matrix(matrix): try: inverse = np.linalg.inv(matrix) return inverse except np.linalg.LinAlgError: print("Matrix is not invertible.") return None def print_matrix(matrix): for row in matrix: print(' '.join(map(str, row))) def main(): matrix = get_matrix_from_input() print("\nMatrix:") print_matrix(matrix) operation = input("\nChoose an operation (transpose or invert): ").lower() if operation == "transpose": result = transpose_matrix(matrix) print("\nTransposed Matrix:") print_matrix(result) elif operation == "invert": result = invert_matrix(matrix) if result is not None: print("\nInverted Matrix:") print_matrix(result) else: print("Invalid operation. Please choose transpose or invert.") if __name__ == "__main__": main() ``` |
Y a continuación ChatGPT nos brinda información relacionada con el código generado, la operación del programa y la librería adicional que debemos instalar para su funcionamiento.
Here's how the program works: 1. The `get_matrix_from_input` function prompts the user to enter the number of rows and columns in the matrix. It then takes input from the user for each element of the matrix, row by row, and returns the matrix as a NumPy array. 2. The `transpose_matrix` function uses the `np.transpose` function from the NumPy library to transpose the given matrix and returns the transposed matrix. 3. The `invert_matrix` function uses the `np.linalg.inv` function from the NumPy library to calculate the inverse of the given matrix. It handles cases where the matrix is not invertible by catching the `LinAlgError` exception and printing a message. If the matrix is invertible, it returns the inverse matrix. 4. The `print_matrix` function takes a matrix and prints it row by row. 5. In the `main` function, the user is prompted to enter the matrix. The matrix is then printed. 6. The user is prompted to choose an operation: transpose or invert. 7. Depending on the chosen operation, the corresponding function is called, and the result is printed. You can run the program and interact with it in the command line. Provide the matrix elements row-wise, specify the operation (transpose or invert), and the program will display the resulting matrix. Note: This program requires the NumPy library, which you can install by running `pip install numpy` in your terminal or command prompt. I hope this helps! Let me know if you have any further questions. |
La respuesta de Bard:
import numpy as np def transpose(matrix): """Transposes a matrix. Args: matrix: A 2D NumPy array. Returns: A 2D NumPy array representing the transposed matrix. """ return matrix.T def invert(matrix): """Inverts a matrix. Args: matrix: A 2D NumPy array. Returns: A 2D NumPy array representing the inverse of the matrix. """ return np.linalg.inv(matrix) def main(): """The main function of the program. This function prompts the user to enter a matrix, and then asks the user if they want to transpose or invert it. The result is then printed to the console. """ # Prompt the user to enter a matrix. matrix_string = input("Enter a matrix: ") # Parse the matrix string into a NumPy array. matrix = np.array(matrix_string.split(), dtype=int) # Prompt the user to choose whether to transpose or invert the matrix. operation = input("Do you want to transpose (t) or invert (i) the matrix? ") # Perform the requested operation on the matrix. if operation == "t": result = transpose(matrix) elif operation == "i": result = invert(matrix) else: print("Invalid operation.") return # Print the result to the console. print("The result is:") print(result) if __name__ == "__main__": main() |
En cuanto a las instrucciones o indicaciones adicionales esto es lo que Bard incluye:
Enter a matrix: 1 2 3 4 5 6 Do you want to transpose (t) or invert (i) the matrix? t The result is: [[1 4] [2 5] [3 6]] |
Conclusión:
En esta ocasión hay un claro ganador, el código de ChatGPT funciona directamente y sin necesidad de hacer nada diferente a copiarlo y seguir las instrucciones en cuanto a la instalación de numpy, por el lado de Bard las funciones centrales del programa se encuentran pendientes por definir.
ChatGPT: 2 Bard: 0
Caso de Uso 3:
Asistente de programación virtual.
En este caso Bard cuenta con una característica que ChatGPT (en su versión gratuita no) y es la de poder acceder a Internet directamente, por esta razón en el prompt que usaremos para Bard pasaremos como argumento una URL mientras que en el caso de ChatGPT será necesario pasar el texto completo. En ambos casos esta es la página que se desea resumir:
Prompt (Bard):
Create an analytical summary of the following web page: https://www.marketwatch.com/story/wework-stock-tumbles-to-record-low-after-news-of-ceo-departure-33fa434d
Prompt (ChatGPT):
Create an analytical summary of the following text ```texto de la página ```
Sorprendentemente la respuesta de Bard es:
I'm not programmed to assist with that. |
Por otra parte ChatGPT devuelve lo siguiente:
WeWork Inc.'s stock experienced a significant decline, reaching a record low after the announcement of its CEO's departure. The stock fell by 24.7% on Wednesday, closing at just 26 cents. Mizuho, an analyst firm, downgraded the stock from a buy rating to a neutral rating and reduced its price target from $1.75 to 30 cents. The sudden departure of CEO Sandeep Mathrani, effective May 26, raised concerns among analysts. Mizuho noted that the abruptness of the decision suggested that it may not have been entirely voluntary. The analyst expects the leadership change to be disruptive, particularly given the current macroeconomic challenges. Mizuho further expressed skepticism about WeWork's ability to achieve its base case business assumptions, particularly occupancy targets. The firm anticipates higher cash burn and the eventual need for external capital, projecting that free cash flow will not turn positive until the end of 2025. This latest blow to WeWork adds to the company's troubled history. Founded by Adam Neumann, WeWork was once valued at $47 billion but faced setbacks, including a botched IPO attempt in 2019 and Neumann's subsequent ousting. Its valuation had dwindled to $9 billion by the time it merged with SPAC BowX Acquisition Corp. Under Mathrani's leadership, WeWork made progress in reducing costs and improving its occupancy rate, which increased from the mid-40s during the COVID-19 pandemic to 75%. The company also restructured its balance sheet, eliminating $1.2 billion in debt. Mizuho acknowledges the potential of WeWork's flexible office space model, considering it as a quasi "call option" if the office sector partially recovers. Their Bull Case scenario assumes occupancy reaching over 80% in the next 12 months, while the Bear Case incorporates corporate restructuring. In its most recent earnings report, WeWork narrowed its net loss for the first quarter of the year and achieved 11% revenue growth. However, the stock's decline and the CEO's departure indicate significant challenges ahead for the company. |
Modificaremos un poco el prompt para darle a Bard una segunda oportunidad:
Prompt (Bard):
Create a summary of the following web page: https://www.marketwatch.com/story/wework-stock-tumbles-to-record-low-after-news-of-ceo-departure-33fa434d
Prompt (ChatGPT):
Create summary of the following text ```texto de la página ```
En este caso Bard modifica su respuesta:
WeWork's stock price fell to a record low on Wednesday after the office-sharing company announced that its chief executive, Sandeep Mathrani, would be stepping down. The stock closed at 26 cents per share, down 24.7% from the previous day. Mathrani's departure comes after a tumultuous period for WeWork, which was forced to postpone its initial public offering last year amid concerns about its corporate governance and financials. The company has since been working to address those concerns, but it remains to be seen whether it will be able to regain investor confidence. |
ChatGPT devuelve lo siguiente:
WeWork Inc., the office-sharing company, experienced a significant drop in its stock price, reaching a record low, following the announcement of its CEO's departure. The stock fell by 24.7% to 26 cents, a stark contrast to its closing price of $11.78 on its first day of trading after going public in October 2021. Mizuho analyst Vikram Malhotra downgraded the stock from a "buy" to a "neutral" rating and reduced the price target to 30 cents from $1.75. The sudden departure of CEO Sandeep Mathrani, effective May 26, raised concerns among analysts, with the downgrade reflecting skepticism about the circumstances surrounding the change. The analyst expects the move to be disruptive, especially considering the current macroeconomic challenges. The analyst also noted that the company's base case business assumptions, particularly occupancy targets, are now seen as unachievable, resulting in increased cash burn and a potential need for external capital. Positive free cash flow is not expected until the end of 2025. WeWork's struggles continue to accumulate, following the departure of its infamous founder, Adam Neumann, in 2019, and a failed initial public offering attempt. The company, once valued at $47 billion, merged with a special-purpose acquisition vehicle (SPAC) at a valuation of $9 billion. Under Mathrani's leadership, WeWork has made progress in reducing costs and improving its occupancy rate, but the CEO change is expected to impact the company's strategy and possibly lead to employee turnover. Despite the challenges, Mizuho maintains some optimism in the flexible office space model and considers WeWork as a quasi "call option" for the sector's partial recovery. The analyst's bullish case envisions driving occupancy to over 80% within the next year, while the bearish case anticipates corporate restructuring. In its most recent earnings report, WeWork narrowed its net loss for the first quarter of the year and achieved revenue growth of 11%. However, the stock's plummet to a record low reflects investor concerns about the company's future prospects in the competitive office-sharing market. |
Finalmente solicitemos a los dos competidores una tabla con los datos clave del resumen:
Prompt (Bard):
Create a summary of the following web page and present the most relevant data in a table: https://www.marketwatch.com/story/wework-stock-tumbles-to-record-low-after-news-of-ceo-departure-33fa434d
Prompt (ChatGPT):
Create summary of the following text and present the most relevant data in a table ```texto de la página ```
Una vez más Bard no sabe qué hacer con esto:
I'm unable to help, as I am only a language model and don't have the ability to process and understand that. |
Por su parte ChatGPT:
Summary: WeWork Inc.'s stock plummeted to a record low of 26 cents following the announcement of its CEO's departure. Mizuho analyst Vikram Malhotra downgraded the stock to neutral from buy and lowered the price target to 30 cents from $1.75. The abruptness of the CEO's departure and the challenging macroeconomic environment were cited as factors contributing to the downgrade. WeWork, once valued at $47 billion, has faced significant setbacks in the past, including the botched IPO and the ousting of its founder Adam Neumann. Under the leadership of the outgoing CEO, the company made progress in cost reduction and debt restructuring. Mizuho sees potential in WeWork's flexible office space model, but the CEO change is expected to bring disruptions and potential employee turnover. Table:  |
Conclusión:
De forma sorprendente, una de las ventajas más claras de Bard (su acceso a Internet) se vuelve su talón de Aquiles, ya que ha fallado en una tarea que la versión gratuita de ChatGPT puede realizar sin inconveniente si se le suministra el texto. Algo que combinando Python con ChatGPT no tiene dificultad.
ChatGPT: 3 Bard: 0
Caso de Uso 4:
Diseño de rutas formativas.
Para finalizar intentaremos diseñar una ruta formativa que incluya los Resultados de Aprendizaje Esperados, la estructura de las sesiones, la evaluación de aprendizajes, las metodologías de enseñanza y la bibliografía.
Prompt:
You are a negotiation expert. Design a training path for participants to learn how to generate win/win agreements. Start the design by defining the Learning Outcomes, continue with the structure of the required sessions, the assessment of learning in each of them, and the suggested teaching methodology for this path. Include the section for suggested bibliography at the end.
¿Qué nos dice Bard?:
**Learning Outcomes** By the end of this training path, participants will be able to: * Understand the principles of win-win negotiation. * Identify their own interests and those of the other party. * Develop creative solutions that meet the needs of both parties. * Communicate effectively during a negotiation. * Build rapport with the other party. * Manage conflict effectively. * Close a negotiation successfully. **Structure of the Required Sessions** The training path will consist of four sessions: * Session 1: Introduction to Win-Win Negotiation * Session 2: Identifying Interests * Session 3: Developing Creative Solutions * Session 4: Closing the Deal Each session will include a combination of lectures, group exercises, and role-playing. **Assessment of Learning** Participants will be assessed on their learning throughout the training path through a variety of methods, including: * Quizzes and exams * Group exercises * Role-playing * Midterm and final projects **Suggested Teaching Methodology** The training path will be taught using a variety of methods, including: * Lectures * Group exercises * Role-playing * Case studies * Discussions The instructor will use a variety of teaching styles, including: * Socratic questioning * Experiential learning * Active learning **Suggested Bibliography** * Fisher, Roger, William Ury, and Bruce Patton. Getting to Yes: Negotiating Agreement Without Giving In. Penguin Books, 2011. * Bazerman, Max H., and Margaret A. Neale. Negotiating Rationally. Free Press, 1992. * Lewicki, Roy J., David M. Saunders, and John W. Minton. Negotiation. McGraw-Hill Education, 2014. * Thompson, Leigh L. The Mind and Heart of the Negotiator. Prentice Hall, 2011. |
¿Cómo responde ChatGPT?:
Training Path: Generating Win/Win Agreements Learning Outcomes: 1. Understand the principles and importance of win/win negotiation. 2. Develop effective communication and active listening skills for negotiation. 3. Identify and manage different negotiation styles and strategies. 4. Apply problem-solving techniques to reach mutually beneficial agreements. 5. Build collaborative relationships and trust with negotiation partners. 6. Analyze and manage power dynamics in negotiations. 7. Practice generating win/win agreements through realistic negotiation simulations. 8. Reflect on personal negotiation strengths and areas for improvement. Structure of Required Sessions: Session 1: Introduction to Win/Win Negotiation - Learning Outcome 1: Understand the principles and importance of win/win negotiation. - Teaching Methodology: Presentation, case studies, group discussions. - Assessment: Group discussion participation, comprehension questions. Session 2: Effective Communication and Active Listening in Negotiation - Learning Outcome 2: Develop effective communication and active listening skills for negotiation. - Teaching Methodology: Role-playing exercises, video examples, group feedback. - Assessment: Role-play performance, active participation in discussions. Session 3: Negotiation Styles and Strategies - Learning Outcome 3: Identify and manage different negotiation styles and strategies. - Teaching Methodology: Lecture, case studies, group activities. - Assessment: Group activities, case study analysis, short quiz. Session 4: Problem-Solving Techniques in Negotiation - Learning Outcome 4: Apply problem-solving techniques to reach mutually beneficial agreements. - Teaching Methodology: Interactive exercises, real-life examples, small group discussions. - Assessment: Group activity participation, analysis of negotiation scenarios. Session 5: Building Collaborative Relationships and Trust - Learning Outcome 5: Build collaborative relationships and trust with negotiation partners. - Teaching Methodology: Role-playing exercises, group discussions, trust-building activities. - Assessment: Role-play performance, active participation in trust-building activities. Session 6: Power Dynamics in Negotiation - Learning Outcome 6: Analyze and manage power dynamics in negotiations. - Teaching Methodology: Case studies, role-playing exercises, group discussions. - Assessment: Group activity participation, analysis of power dynamics in negotiation scenarios. Session 7: Negotiation Simulations - Learning Outcome 7: Practice generating win/win agreements through realistic negotiation simulations. - Teaching Methodology: Simulated negotiations, debriefings, feedback sessions. - Assessment: Negotiation simulation performance, reflection papers. Session 8: Reflection and Personal Development - Learning Outcome 8: Reflect on personal negotiation strengths and areas for improvement. - Teaching Methodology: Self-assessment activities, guided reflection, goal setting. - Assessment: Personal reflection papers, goal setting exercise. Suggested Bibliography: 1. "Getting to Yes: Negotiating Agreement Without Giving In" by Roger Fisher and William Ury. 2. "Negotiation Genius: How to Overcome Obstacles and Achieve Brilliant Results at the Bargaining Table and Beyond" by Deepak Malhotra and Max Bazerman. 3. "Getting Past No: Negotiating in Difficult Situations" by William Ury. 4. "Difficult Conversations: How to Discuss What Matters Most" by Douglas Stone, Bruce Patton, and Sheila Heen. 5. "Bargaining for Advantage: Negotiation Strategies for Reasonable People" by G. Richard Shell. 6. "Negotiating Rationally" by Max H. Bazerman and Margaret A. Neale. 7. "Crucial Conversations: Tools for Talking When Stakes Are High" by Kerry Patterson, Joseph Grenny, Ron McMillan, and Al Switzler. 8. "Never Split the Difference: Negotiating As If Your Life Depended On It" by Chris Voss and Tahl Raz. Note: The suggested bibliography includes books that cover various aspects of negotiation and communication skills, providing participants with a comprehensive understanding of win/win negotiation. The training path incorporates a mix of teaching methodologies to engage participants actively and promote practical application of the concepts learned. Assessment methods involve individual and group activities, simulations, quizzes, and reflective writing to assess participants' comprehension and skill development throughout the training path. |
Analicemos los resultados de acuerdo con los criterios solicitados en el prompt.
Comparemos las respuestas:
Criterio | Bard | ChatGPT | Resultado |
Resultados de Aprendizaje ILOs | Genera 7 ILOs de acuerdo con lo solicitado. | Genera 8 ILOs de acuerdo con lo solicitado. | Los dos han generado lo solicitado de acuerdo con la estructura esperada de un Resultado de Aprendizaje. |
Estructura de las sesiones requeridas | Lista las 4 sesiones del curso con el tema central a desarrollar y señala de manera general aspectos relacionados con la evaluación y la metodología de enseñanza. | Presenta una estructura de 8 sesiones para cada una de las cuales asocia el Resultado de Aprendizaje correspondiente, la evaluación de aprendizajes indicada y la metodología de enseñanza | La propuesta de ChatGPT es más completa pero además promueve la coherencia curricular al garantizar la alineación de cada sesión con los ILOs definidos y de la evaluación con el ILO específico de cada sesión. |
Bibliografía | Incluye 4 libros los cuales existen (no son fuentes falsas). | Incluye 8 libros los cuales existen (no son fuentes falsas). Adicionalmente propone títulos que no hacen parte de un curso tradicional de negociación pero que lo hacen inmensamente valioso, como el caso del libro Never Split the Difference de Chriss Voss. | ChatGPT sugiere una bibliografía más completa y considera enfoques no tradicionales dentro de la negociación integrativa. |
Conclusión:
Una vez más ChatGPT ha resultado vencedor
ChatGPT: 4 Bard: 0
Se han puesto a prueba los LLM de Open AI y Google y en cuatro (4) casos de uso específicos ChatGPT ha resultado vencedor. Evidentemente puede ser un tema de madurez, el producto de OpenAI lleva más tiempo en funcionamiento y con cada versión se ve como mejora de forma incremental (las 8 referencias verdaderas en el último caso de uso era algo impensable en el release que se encontraba disponible en enero o febrero de este año. Sin embargo, no deja de llamar la atención el bajo desempeño de Bard en dos tareas en las que se esperaría fuera superior: resumiendo información de páginas web a las que puede acceder directamente y generando códigos de programación.
La competencia entre Bard y ChatGPT nos deja con una sensación de emoción y curiosidad, deseosos de presenciar cómo estos gigantes de la generación de lenguaje seguirán evolucionando y desafiándose según el futuro. ¡La puerta hacia nuevas posibilidades en la era de la inteligencia artificial está abierta para aquellos que deseen seguir explorando los límites de Bard y ChatGPT!