9 of 105

Definitions

ChatGPT is a chatbot fine-tuned to produce human-like responses to prompts.

ChatGPT is powered by a large language model, trained on massive datasets, that predicts the next word of its output from previous words.

ChatGPT

Natural Language Processing

Neural Network Architecture

Icons from thenounproject

10 of 105

GPT versions

GPT-3.5 is an older model, and is free

GPT-4 is newer, costs $20/month, and OpenAI limits messages to 25 per 3 hours. However, it has superior performance

ChatGPT

Natural Language Processing

Neural Network Architecture

Icons from thenounproject

11 of 105

Other generative AIs

Photo Illustration by Pavlo Gonchar/SOPA Images/LightRocket via Getty Images

Futurum

*Bing uses GPT-4 with modifications (Microsoft is an investor in OpenAI)

12 of 105

ChatGPT excels at producing human-like text in many forms and styles

Your task is to write an NIH specific aims page about sensory processing in electric fish. Use academic language and a formal tone.

→ This isn’t an existing aims page ChatGPT pulled off the internet; it’s new (full text)

Icons from thenounproject

13 of 105

ChatGPT excels at producing human-like text in many forms and styles

Project Title: Investigation of the Neural Mechanisms Underlying Sensory Processing in Electric Fish

Introduction: Electric fish are a unique group of aquatic animals that generate electric fields to navigate their environment and communicate with other individuals. They possess a specialized sensory system that enables them to detect, encode, and interpret the electrical signals produced by their own and other individuals' electric organs. Understanding how electric fish process sensory information is of fundamental importance for advancing our knowledge of the neural basis of perception and cognition.

Specific Aims:

Aim 1: Characterize the neural circuits involved in sensory processing in electric fish

To achieve this aim, we will use a combination of in vivo electrophysiology, optogenetics, and neuroanatomical techniques to identify and characterize the neural circuits that underlie the processing of electric signals in the brain of electric fish. We will record neural activity from the brain of electric fish while presenting them with different types of electric stimuli and use optogenetics to selectively activate or inhibit specific neurons or circuits to determine their functional roles in sensory processing.

full text

ChatGPT

14 of 105

ChatGPT’s training data don’t include recent events

As an AI language model, I cannot provide real-time information or updates on events that have occurred after my knowledge cutoff date in September 2021. Therefore, I cannot identify the most important scientific breakthrough of 2022. However, up until my last update, there were several ongoing research areas that had the potential for significant breakthroughs, such as the development of new vaccines, advancements in renewable energy technologies, artificial intelligence, and space exploration, among others. Please consult more recent sources to get accurate information on the most important scientific breakthroughs of 2022.

What was the most important scientific breakthrough of 2022?

Icons from thenounproject

15 of 105

ChatGPT invents information (“hallucinates”)

As of my last knowledge update in September 2021, Tim Requarth is an American freelance journalist, writer, and lecturer. He has written for various publications, including The New York Times, Slate, and The Nation, among others. Requarth has covered a range of topics, but he often focuses on science, medicine, and health. In addition to his journalism career, he has experience teaching science writing and has held various teaching and fellowship positions at institutions such as Columbia University. Please note that this information may be outdated, and I recommend checking the latest sources for any updates on his work or career.

As of my knowledge cutoff date in September 2021, Tim Requarth has written various articles on science, health, and environmental topics. Some of his articles include:

"Why Do People Fall for Fake News?" - published in The New York Times on January 18, 2017. In this article, Requarth explores the psychological reasons behind people's susceptibility to fake news and misinformation.

"This Is Your Brain on Drugs" - published in Slate on April 17, 2014. This article examines how drugs interact with the brain and the complex factors that determine whether a person becomes addicted to a substance.

Who is tim requarth?

What articles has he written?

Hallucinations

Icons from thenounproject

16 of 105

ChatGPT invents information (“hallucinates”)

Title: First Republic Bank Surpasses JPMorgan in Market Capitalization

In a surprising turn of events, First Republic Bank has overtaken JPMorgan Chase & Co. in terms of market capitalization, according to a recent article published by The New York Times. The California-based bank has experienced rapid growth in recent years, thanks to its focus on providing personalized services to high-net-worth individuals and businesses. As a result, its market capitalization has reached new heights, surpassing that of the once-dominant JPMorgan. The article goes on to discuss the factors that have contributed to First Republic Bank's success, as well as the potential implications for the banking industry at large.

Summarize the content of this web page: https://www.nytimes.com/2023/05/01/business/first-republic-bank-jpmorgan.html

Hallucinations

Icons from thenounproject

???

17 of 105

Summary of limitations

The training data is only current through Sep 2021

The training data only partially includes the primary scientific literature

Generative AI tools like ChatGPT have a tendency to “hallucinate”

Output replicates biases and inaccuracies of training data

Generative AI tools are excellent pattern-matchers but limited reasoners

→ Some are harder to address

→ Some of these limitations will be quickly overcome

18 of 105

Disclaimer: Using AI tools can violate journal policies

Text generated from AI tools cannot be used at all

Text generated from AI tools is “discouraged” but must be disclosed with details

Science Journals: Editorial Policies

Instructions for Authors | Nature

Instructions for Authors | JAMA

AI tools using LLMs cannot be authors, and must be disclosed in Methods

19 of 105

AI is being integrated directly into productivity apps

Microsoft copilot

21 of 105

Use case: Revising a passage for grammar

Rewrite the text using correct grammar and punctuation: Facebook, with more then 1 Billion users Is celebrated its 10th birthday this year. The company was launching by Mark Zuckerberg at 4th February 2004. The site was Conceived in a Harvard dorm in a way To connect student and let them build an Identity for himself online.

Facebook, with more than 1 billion users, is celebrating its 10th birthday this year. The company was launched by Mark Zuckerberg on February 4th, 2004. The site was conceived in a Harvard dorm as a way to connect students and let them build an identity for themselves online.

Example from leverage edu

Icons from thenounproject

→ This is fantastic if English isn’t your primary language

22 of 105

Use case: Revising a passage for clarity

Arc is an essential gene expressed in neurons, critical for long-lasting storage of information in the brains of mammals. Arc, which has been implicated in various neurodevelopmental disorders, also plays a known role in the mediation of a number of forms of synaptic plasticity. However, in spite of the importance of Arc, a full understanding of its molecular function has not been elucidated. In addition, the evolutionary origins of Arc are not very well understood. Here, we provide evidence for the self-assembly of Arc into capsids that resemble viruses, including the encapsulation of nucleic acids—in this case, RNA. We go on to show that the transfer of Arc mRNA into new target cells is mediated by endogenous Arc protein being released from neurons in extracellular vesicles. In the new target cells, we show that the end result of this process is an increase in the translation of Arc protein when there is an elevation of neuronal activity. In terms of its molecular properties, we conclude that Arc has a resemblance to retroviral proteins belonging to the Gag family. In terms of evolution, our analysis suggests it is highly likely that Arc is derived from a lineage of Ty3/gypsy retrotransposons, which are found in vertebrates, and which are also ancestors to retroviruses. Thus, our experiments and analysis have led to the identification of a Gag retroelement that appears to have taken on a new role during the course of evolution mediating communication among neurons in the nervous system.

Adapted from Pastuzyn et al Cell 2018

23 of 105

Use case: Revising a passage for clarity

Arc is a crucial gene expressed in neurons that plays a vital role in the long-term storage of information in mammalian brains. Despite its significance, there is still much to be understood about its molecular function and evolutionary origins. Our recent study provides evidence of Arc's self-assembly into capsids resembling viruses, which can encapsulate RNA. Furthermore, we demonstrate that endogenous Arc protein released from neurons in extracellular vesicles facilitates the transfer of Arc mRNA into new target cells. This results in an increase in Arc protein translation when there is an escalation of neuronal activity. Our analysis indicates that Arc has similarities to retroviral proteins in the Gag family and is most likely derived from the lineage of Ty3/gypsy retrotransposons, which are ancestors of retroviruses and found in vertebrates. Through our findings, we have identified a Gag retroelement that has evolved to play a crucial role in facilitating communication among neurons in the nervous system.

Revise the following text for clarity: “YOUR TEXT”

Adapted from Pastuzyn et al Cell 2018

Icons from thenounproject

24 of 105

ChatGPT can’t always identify logical gaps

The enthalpy of hydrogen bond formation between the nucleoside bases 2’deoxyguanosine (dG) and 2’deoxycytidine (dC) has been determined by direct measurement. dG and dC were derivatized at the 5’ and 3’ hydroxyls with triisopropylsilyl groups to obtain solubility of the nucleosides in non-aqueous solvents and to prevent the ribose hydroxyls from forming hydrogen bonds. From isoperibolic titration measurements, the enthalpy of dC:dG base pair formation is -6.65+0.32 kcal/mol.

Example from Gopen and Swan 1990

25 of 105

ChatGPT can’t always identify logical gaps

The enthalpy of hydrogen bond formation between the nucleoside bases 2'deoxyguanosine (dG) and 2'deoxycytidine (dC) has been determined through direct measurement. To obtain solubility of the nucleosides in non-aqueous solvents and prevent the ribose hydroxyls from forming hydrogen bonds, dG and dC were derivatized at the 5' and 3' hydroxyls using triisopropylsilyl groups. Isoperibolic titration measurements revealed that the enthalpy of dC:dG base pair formation is -6.65 ± 0.32 kcal/mol.

We have directly measured the enthalpy of hydrogen bond formation between the nucleoside bases 2’deoxyguanosine (dG) and 2’deoxycytidine (dC). dG and dC were derivatized at the 5’ and 3’ hydroxyls with triisopropylsilyl groups; these groups serve both to solubilize the nucleosides in non-aqueous solvents and to prevent the ribose hydroxyls from forming hydrogen bonds. Consequently, when the derivatized nucleosides are dissolved in non-aqueous solvents, hydrogen bonds form almost exclusively between the bases. Since the interbase hydrogen bonds are the only bonds to form upon mixing, their enthalpy of formation can be determined directly by measuring the enthalpy of mixing. From our isoperibolic titration measurements, the enthalpy of dC:dG base pair formation is -6.65+0.32 kcal/mol.

Revise the following text: The enthalpy of hydrogen bond formation between the nucleoside bases 2’deoxyguanosine (dG) and 2’deoxycytidine (dC) has been determined by direct measurement. dG and dC were derivatized at the 5’ and 3’ hydroxyls with triisopropylsilyl groups to obtain solubility of the nucleosides in non-aqueous solvents and to prevent the ribose hydroxyls from forming hydrogen bonds. From isoperibolic titration measurements, the enthalpy of dC:dG base pair formation is -6.65+0.32 kcal/mol.

Example from Gopen and Swan 1990

Logical gap

human

Icons from thenounproject

26 of 105

Use case: Applicant statement

Your task is to write a personal statement for an application for an NIH F31 NRSA fellowship. Include three detailed research experiences. The personal statement should demonstrate how these experiences have prepared the applicant to be successful in pursuing the NIH F31 NRSA fellowship. You should focus on providing clear and concise details about both topics while still maintaining a formal tone throughout.

Icons from thenounproject

27 of 105

Use case: Applicant statement

Pretty generic

Good template, wrong content

28 of 105

Use case: Applicant statement

This output is only useful for structure, not content

29 of 105

Use case: Applicant statement research experience paragraph

What motivated you to seek out this experience?

When/where did this experience happen?

What was the project?

What was your role in it?

What were the results or any other tangible product of your experiences?

Did you receive any awards or fellowships to support your work? How did you communicate your results (paper, poster, etc)?

What skills or lessons did you learn from the experience and how did they affect your motivation or career trajectory?

Template for writing about a research experience

template source

Icons from thenounproject

30 of 105

Answer with rough writing or voice-to-text

Photo by Aaron Burden

voice-to-text

ChatGPT

Organize info, extract themes, etc

31 of 105

Browser extensions can streamline the process

Chrome extension called “GPT for Sheets and Docs”

32 of 105

Use case: Applicant statement (voice-to-text)

Use case: Writing an introduction section

The ability to generate negative images of the sensory consequences of one's own movements is crucial for animals to accurately perceive their external environment. The neural mechanisms underlying this ability are well-studied in some systems, such as the cerebellum in mammals. However, the mechanisms for generating negative images in other systems, such as the electrosensory system in weakly electric fish, remain poorly understood. In this study, we investigated the generation of negative images in the electrosensory system of the weakly electric fish, using a combination of in vivo electrophysiology and microstimulation techniques. We found that a subset of mossy fibers in the cerebellum analogue of the fish, the eminentia granularis posterior (EGp), convey graded motor information related to the frequency of rhythmic swimming movements, and that these fibers exhibit firing rate modulation that is correlated with motor nerve activity. Further experimentation revealed that corollary discharge signals related to movements are conveyed via separate pathways from electromotor command signals, and that these corollary discharge inputs to EGp neurons are plastic, being shaped by mechanisms of associative plasticity. Moreover, we found that ELL neurons, which receive inputs from EGp and are responsible for generating negative images of the electrosensory consequences of movements, possess the capacity to transform motor commands into much longer-lasting patterns of activity that are temporally-aligned with and appropriate to cancel the electrosensory consequences of movements. Our data also suggest that the generation of negative images in the electrosensory system is highly flexible and robust, and that negative images based on both corollary discharge and proprioception require spinal input. These findings provide new insights into the neural mechanisms underlying the generation of negative images in the electrosensory system, and may have broader implications for understanding how animals form accurate perceptions of their external environment.

39 of 105

Use case: Writing an introduction section

Not so great - minimal background, doesn’t set up problem, mostly summarizes results

40 of 105

The ‘CARS’ (Create a Research Space) model

Establish the research space

Make the research feel inevitable and important (the ‘so-what’ factor)

Establish the niche

Define the gaps in knowledge
Define any barriers to progress
State any critical needs

Occupy the niche

Explain how the research fills the gaps in knowledge, overcomes the barriers to progress, or meets the critical needs

CARS model

Icons from thenounproject

41 of 105

Use case: Writing an introduction section

I'm going to give you an abstract and results section, and I want you to provide a detailed template for an introduction in the style of the CARS (‘Create a Research Space’) model for the introductions to scientific papers. Here's the text: “YOUR TEXT”

Icons from thenounproject

42 of 105

Use case: Writing an introduction section

Establishing a territory:

Begin by discussing the importance of predicting sensory consequences of movements in various aspects of an organism's function, such as sensory processing, motor control, and cognitive processes.
Mention previous research that has investigated the role of corollary discharge in generating predictions about the sensory consequences of movements.

Establishing a niche:

Point out the limitations of current knowledge in understanding how corollary discharge operates at the neural circuit level.
Discuss the significance of studying the neural mechanisms underlying the generation of negative images of sensory consequences in organisms, such as weakly electric mormyrid fish.

Occupying the niche:

State the main objective of the study, which is to investigate how neurons in the electrosensory lobe (ELL) of weakly electric mormyrid fish generate negative images of sensory consequences based on ascending spinal corollary discharge signals.
Briefly describe the methodology used to demonstrate this neural mechanism in mormyrid fish.
Mention that the findings of this study not only generalize previous results about the generation of negative images in relation to the fish's electric organ discharge (EOD) but also suggest the involvement of a cerebellum-like circuit with associative synaptic plasticity.
Conclude by emphasizing the potential implications of these findings for understanding how the brain solves the complex problem of predicting sensory consequences of movements across various species.

43 of 105

Use case: Writing an introduction section

Provide a more detailed template of 1 (establishing the territory) and 2 (establishing the niche).

Icons from thenounproject

44 of 105

Use case: Writing an introduction section

Establishing the territory:

1.1. Introduce the importance of sensory and motor systems in various aspects of an organism's function, such as perception, action, and cognition.
1.2. Explain the concept of sensory predictions and their role in facilitating efficient processing of sensory information and guiding motor actions.
1.3. Discuss the prevailing hypothesis that internal signals related to motor commands, known as corollary discharge, serve to generate predictions of sensory consequences of movements.
1.4. Provide a brief overview of previous research on corollary discharge, its functional role, and the neural basis of this process.
1.5. Highlight the unique features of the weakly electric mormyrid fish's electrosensory system, making it an interesting model for studying sensory predictions and corollary discharge.

Establishing the niche:

2.1. Point out that despite the existing body of research on corollary discharge, its neural circuit-level mechanisms remain poorly understood.
2.2. Emphasize the need for further investigation into the neural underpinnings of corollary discharge and its role in generating negative images of sensory consequences.
2.3. Discuss the potential benefits of understanding these mechanisms in weakly electric mormyrid fish, as their specialized electric organ discharge (EOD) system can offer valuable insights into the general principles of sensory predictions and adaptive filtering in other organisms.

45 of 105

Use case: Writing an introduction section

Establishing the territory:

1.1. Introduce the importance of sensory and motor systems in various aspects of an organism's function, such as perception, action, and cognition.
1.2. Explain the concept of sensory predictions and their role in facilitating efficient processing of sensory information and guiding motor actions.
1.3. Discuss the prevailing hypothesis that internal signals related to motor commands, known as corollary discharge, serve to generate predictions of sensory consequences of movements.
1.4. Provide a brief overview of previous research on corollary discharge, its functional role, and the neural basis of this process.
1.5. Highlight the unique features of the weakly electric mormyrid fish's electrosensory system, making it an interesting model for studying sensory predictions and corollary discharge.
2.3. Discuss the potential benefits of understanding these mechanisms in weakly electric mormyrid fish, as their specialized electric organ discharge (EOD) system can offer valuable insights into the general principles of sensory predictions and adaptive filtering in other organisms.

Establishing the niche:

2.1. Point out that despite the existing body of research on corollary discharge, its neural circuit-level mechanisms remain poorly understood.
2.2. Emphasize the need for further investigation into the neural underpinnings of corollary discharge and its role in generating negative images of sensory consequences.

My rearrangement

46 of 105

Use case: Writing a discussion section

Generate ideas for a discussion section for this paper: “YOUR TEXT”

Requarth & Sawtell, Neuron (2014)

47 of 105

Use case: Writing a discussion section

Significance of the findings: Emphasize the importance of the discovery that ELL neurons receive movement-related corollary discharge signals and can generate negative images of the electrosensory consequences of different movements. Discuss how this extends our understanding of the ELL's role in processing electrosensory information and motor control.

Integration of corollary discharge signals with other sensory inputs: Discuss how the integration of movement-related corollary discharge signals with electrosensory and proprioceptive inputs contributes to the generation of a comprehensive representation of the animal's movements and the surrounding environment. This could potentially enhance the fish's ability to navigate, avoid predators, and locate prey.

Comparison with other systems: Compare the findings in this study with the known functions of corollary discharge signals in other animals and sensory systems. Explore the similarities and differences in how different species utilize corollary discharge signals for sensory processing and motor control.

Neural plasticity and adaptive function: Highlight the role of neural plasticity in shaping the corollary discharge inputs to ELL neurons. Discuss how the observed plasticity may contribute to the generation of multiple negative images related to different movements, allowing the fish to adapt to various movement patterns and sensory contexts.

Potential circuit mechanisms: Explore the potential circuit mechanisms underlying the generation of negative images in ELL neurons based on corollary discharge signals. Discuss how the observed plasticity at parallel fiber-MG cell synapses might contribute to this process.

Limitations and challenges: Address any limitations or challenges encountered in the study, and discuss how these might be overcome in future research. For example, consider improvements to the experimental design or additional control experiments that could provide further insights into the role of corollary discharge signals in ELL neurons.

Future research directions: Suggest avenues for future research, such as investigating the role of other brain regions in processing and integrating corollary discharge signals, studying the neural mechanisms of negative image formation in more detail, and examining the potential behavioral consequences of the observed plasticity in ELL neurons. Additionally, propose potential experiments to further elucidate the role of corollary discharge signals in electrosensory and motor processing in other species.

Requarth & Sawtell, Neuron (2014)

These ideas must be vetted carefully

48 of 105

Use case: Writing a discussion section

What differentiates your study from previous studies? What has been added to knowledge, technology, or methodology by your work?
How consistent or inconsistent are your results with prior studies?
Are there alternative explanations for any of your main findings?
Are there any mechanisms or other topics worth speculating about?
What are the limitations of your study? What questions to these limitations leave unanswered or ambiguous? What are the key experiments or analyses that must be done?
What is the “take-home” message you want your reader to leave with?
What impact could your work ultimately have on medicine, technology, or society?

Adapted from How to write discussions and conclusions (PloS)

Questions to ask yourself:

49 of 105

Use case: Writing a discussion section

Source: Discussion section with ChatGPT

Icons from thenounproject

50 of 105

Use case: Writing a discussion section

I answered a series of questions that served as prompts for me to get started writing a discussion section in a scientific manuscript. I also have a discussion section structure that I'd like to use. I'm going to give you that structure and then give you my answers, and I want you to draft a discussion section. Each paragraph should be at least 4 sentences, and I want you to retain detail and include analysis.

Here is the structure:

Paragraph 1: Restate key findings

Paragraphs 2-4: Pick a few main findings and in one paragraph per finding discuss one or more of the following aspect(s):

Compare/contrast to previous studies

Highlight unexpected finding

Strengths/limitations

Possible mechanisms

Unanswered questions/future research

Paragraph 5: Discuss broader implications and future research

Here are the questions followed by each answer: “YOUR TEXT”

Icons from thenounproject

Source: Discussion section with ChatGPT

51 of 105

Use case: Writing a discussion section

Paragraph 1: In our study, we extended the investigation of corollary discharge in electric fish from the highly specialized electro motor system to the more general musculoskeletal system, specifically focusing on tail movements. We discovered that corollary discharge ascends from the spinocerebellar tract, which is consistent with previous findings in mammals. This finding is significant because it expands our understanding of corollary discharge beyond the electric organ and into the broader realm of musculoskeletal movements. Furthermore, our work highlights the importance of investigating the cerebellum-like circuit in electric fish for understanding the general principles of sensory filtering and motor control in other species.

Writing is clear but a little flat, stylistically

Source: Discussion section with ChatGPT

52 of 105

Use case: Writing a discussion section

Paragraph 2: Our findings build on previous studies that focused on the stereotyped electric organ discharge, providing insights into the more complex and diverse musculoskeletal movements. The discovery of corollary discharge from the spinocerebellar tract in fish is novel but consistent with similar findings in mammals, suggesting a conserved mechanism across species. This consistency strengthens our understanding of corollary discharge and its role in canceling self-generated sensory input across different organisms and systems. By extending the investigation of corollary discharge to the musculoskeletal system, our study offers a more comprehensive view of the underlying neural mechanisms and their potential relevance to other animals.

No citations

Repetitive, little analysis

Source: Discussion section with ChatGPT

53 of 105

Use case: Writing a discussion section

Paragraph 3: The limitations of our study include the use of electrically stimulated movements, which may not accurately reflect natural movements in the fish. These artificial movements could potentially generate raw material for the cerebellum-like circuit that would not be present in real-life conditions. Furthermore, the electrical stimulation may result in stronger signals than would be naturally generated, leading to an overestimation of the role of corollary discharge in real-life scenarios. Future research could involve conducting experiments using more natural movements or artificially triggered sensory input based on the fish's natural swimming movements to better understand the role of corollary discharge in realistic scenarios. This would help to further validate our findings and improve the generalizability of our results.

Not bad, but needs polishing

Source: Discussion section with ChatGPT

54 of 105

Use case: Writing a discussion section

Paragraph 4: Our study raises questions about the mechanisms by which a more extensive range of movements can be stored in the cerebellum-like circuit of electric fish. We demonstrated that the circuit could store negative images for simple left and right tail movements, but it remains unclear how the circuit manages a richer space of movements. Investigating the roles of different cell types, such as Golgi cells or unipolar brush cells, in this process could help elucidate the underlying neural mechanisms. Moreover, understanding how the granule cells form the basis for this richer space of movements, as opposed to the simpler sensory consequences of the electromotor command, will provide valuable insights into the capacity of the circuit to generate a greater number of negative images. Further research in this area would help to uncover the full capabilities of the cerebellum-like circuit in electric fish and its potential applications to other animals.

No specific experiments suggested

Source: Discussion section with ChatGPT

55 of 105

Use case: Writing a discussion section

Paragraph 5: The broader implications of our study relate to longstanding theories of cerebellar function, which remain largely untested. Our findings suggest that the cerebellum-like circuit in electric fish may have more general applicability to other animals. While the impact of our work on medicine and technology may be distant, it could contribute to our understanding of disorders like schizophrenia and autism, where sensory filtering may play a role. Our work could ultimately influence the development of brain theories and provide insights into the neurobiological mechanisms underlying these disorders. By establishing a link between the cerebellum-like circuit in electric fish and the general principles of sensory filtering and motor control, our study helps to bridge the gap between highly specialized systems and broader neurological processes, offering a foundation for future research in the field.

Clinical relevance is poorly explored

Theories need to be expanded on

Source: Discussion section with ChatGPT

56 of 105

Use case: Journal submission letter

A little generic but provides a starting point

Write a submission letter to the journal Nature for a paper titled "Neural Mechanisms of Sensory Filtering in Mormyrid Fish." The paper abstract is here: “YOUR TEXT”

I hope this letter finds you well. I am writing to submit our original research paper entitled "Neural Mechanisms of Sensory Filtering in Mormyrid Fish" for consideration for publication in Nature. My co-authors and I believe that our study presents novel findings that significantly contribute to the understanding of sensory processing and filtering mechanisms in electrosensory organisms, with potential implications for the broader field of neuroscience.

In this study, we investigated the neural mechanisms underlying sensory filtering in the weakly electric mormyrid fish, which possesses a unique electrosensory system for detecting and processing electrical signals in their environment. Using a combination of electrophysiology, behavioral assays, and computational modeling, we were able to uncover the role of specific neural circuits and their interactions in the context of electrosensory filtering.

Our findings reveal that mormyrid fish employ a highly specialized network of neurons that actively suppress redundant or irrelevant sensory inputs, allowing the fish to focus on more salient information. These results not only enhance our understanding of the neural mechanisms underlying sensory filtering in mormyrid fish, but also provide valuable insights into the general principles of information processing in the nervous systems of other species, including humans.

We believe that the comprehensive approach we have taken in our study, as well as the potential implications of our findings for the broader field of neuroscience, make this paper an excellent fit for publication in Nature. To our knowledge, this is the first study to elucidate the specific neural mechanisms of sensory filtering in mormyrid fish, and we are confident that it will be of great interest to the journal's readership.

We have carefully prepared our manuscript in accordance with Nature's formatting and submission guidelines and have no conflicts of interest to declare. The manuscript has not been submitted to any other journal and is not currently under consideration for publication elsewhere.

Thank you for considering our submission, and we look forward to the opportunity to share our work with the scientific community through Nature. We are confident that our findings will contribute significantly to the ongoing discourse on sensory processing and neural mechanisms. Please do not hesitate to contact me if you require any further information or clarification.

Icons from thenounproject

57 of 105

Use case: Journal submission letter

Totally not true!

Lesson: ChatGPT can provide structure & inspiration, not the final product

…We believe that the comprehensive approach we have taken in our study, as well as the potential implications of our findings for the broader field of neuroscience, make this paper an excellent fit for publication in Nature. To our knowledge, this is the first study to elucidate the specific neural mechanisms of sensory filtering in mormyrid fish, and we are confident that it will be of great interest to the journal's readership…

Icons from thenounproject

58 of 105

Use case: Preparing a general audience summary

Use case: Strengthening arguments

63 of 105

Learn to optimize your prompts (“prompt engineering”)

64 of 105

Learn to optimize your prompts (“prompt engineering”)

write an applicant statement for an NIH NRSA F31 including three research experiences

Please write a statement of purpose for an NIH NRSA F31 research grant application. Your statement should be approximately 500-750 words and focus on your qualifications, experiences, and future goals related to the proposed project. Specifically, please provide detailed descriptions of three relevant research experiences that demonstrate your skills in conducting independent research. For each experience, you should explain how it has prepared you for this project and what specific knowledge or techniques you have gained from it. Additionally, please discuss any other relevant qualifications that make you well-suited for the position. Finally, please include a clear description of your current career objectives and how this project will help you achieve them.

The optimized prompt is more specific than the original prompt by providing clearer guidance on the type of response required and focusing on key elements such as qualifications, experiences, and future goals. It also encourages flexibility and creativity while still maintaining a clear structure and focus on accuracy. The optimized prompt requires the user to describe their relevant research experiences in detail and explain how they have prepared them for the proposed project, as well as discuss any other relevant qualifications that make them suitable for the position. Additionally, it requires them to clearly articulate their current career objectives and how this project will help them achieve those goals.

Promptperfect

Original prompt

Optimized prompt

Explanation

65 of 105

Takeaway: AI tools still require significant human oversight

rough writing or voice-to-text

ChatGPT

first draft

human and AI revisions

Icons from the noun project

Icons from thenounproject

template

66 of 105

Takeaway: AI tools still require significant human oversight

template

rough writing or voice-to-text

ChatGPT

first draft

human and AI revisions

Natural intelligence

Artificial intelligence

more efficient writing

Icons from the noun project

Icons from thenounproject

67 of 105

Final thoughts & additional tools

68 of 105

AI tools can assist with every step

generate a roadmap

write a rough draft

revise

gather ideas & materials

publish

69 of 105

AI tools can assist with every step

generate a roadmap

write a rough draft

revise

gather ideas & materials

publish

Use cases:

-generate discussion ideas

-generate titles

-free writing

Use cases:

-applicant statement

-journal letter

-paper intro

Use cases:

-applicant statement w/template

-journal letter

-paper intro

-results paragraph

-lay summary

Use cases:

-grammar

-clarity

70 of 105

Beware the ‘Tesla effect’

TechCrunch Sep 2021

71 of 105

AI use requires skills at both input and output stages

input (prompts)

ChatGPT

output (filtering)

72 of 105

Potentially useful add-ons for using ChatGPT

→ Most cost money, and you should be mindful of privacy

Writingmate - select a block of text in your browser and ask ChatGPT to respond to an email, rewrite something, etc

Merlin - select a block of text in your browser and ask ChatGPT to respond to an email, rewrite something, etc; also provides side-by-side search results

GPT for Sheets and Docs - use ChatGPT within Google sheets and docs

Promptheus - use your voice to talk to ChatGPT with the spacebar

promptperfect - optimize prompts

73 of 105

Other AI writing aids

Writefull - a grammar and usage checker than can also paraphrase and generate abstracts

Wordtune - Rewrites sentences and passages for clarity and tone, summarizes long articles

Paperpal - A grammar and usage checker tailored to academic writing

→ Most cost money, and you should be mindful of privacy

Grantable - AI writing toolkit catered to grantwriting

74 of 105

non-ChatGPT research tools

Elicit - Helps with literature searches, including finding and extracting info from relevant papers

Scite - Visualize citation networks, get summary of when a finding is supported/refuted in literature, AI chatbot

ConnectedPapers - Visualize citation networks, get prior and derivative works

R discovery - Helps with keeping up with the literature, using AI to generate top recent papers in field, etc

→ Most cost money, and you should be mindful of privacy

75 of 105

Contact & Feedback:

If you aren’t attending the hands-on workshop, please consider leaving feedback on the lecture using the link or QR code below

https://forms.gle/2QufZ8CtjcMyCPz48

Slides & recording will be distributed

With other questions, feel free to email me: timrequarth@gmail.com

76 of 105

Photo by Ana Municio

77 of 105

Workshop: Using AI writing tools in your scientific writing process

Tim Requarth

Lecturer in Science & Writing, Vilcek Institute

NYU Grossman School of Medicine

May 2, 2023

78 of 105

Blank screen anxiety is real

79 of 105

Reflect:

What are your writing roadblocks?

Which parts of the writing process do you dread?

What do you think AI can help you with and what do you think it can’t help you with?

Photo by Paola Aguilar

80 of 105

Agenda

Lecture: Diagnosing writing problems

Activity: Identify the challenges you face with your writing sample

Lecture: Mapping AI strategies to specific writing challenges

Activity: Apply AI assistance to your writing sample

Photo by Mark König

81 of 105

Diagnosing writing problems

82 of 105

Three levels of revising

Conceptual - Deciding what material to include, what the argument is, how big ideas fit together

Photo by Vardan Papikyan

Structural - Ordering material for logical flow and persuasive effect

Stylistic - Tweaking sentences and words for clarity and stylistic convention

Photo by Yana Marudova

Photo by Tim Johnson

Takeaways

Address conceptual and structural problems before stylistic ones

No amount of stylistic tinkering can solve structural or conceptual issues

Photo by Unseen Studio

89 of 105

Activity

Diagnose the problem(s) with your current text. Use the “three levels of revising” or the “quadrants” rubrics as guides if you wish

Photo by National Cancer Institute

90 of 105

Mapping AI strategies to writing challenges

91 of 105

Identify your goal

Concepts

Structure

Text

Icons from thenounproject

92 of 105

Spaces in the writing process

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

93 of 105

Quadrant 1: Clear thinking, too much text

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

-grammar

-clarity

-logical flow

94 of 105

AI strategies: Quadrant 1

Rewrite the text using correct grammar and punctuation: “YOUR TEXT”

Rewrite the text for clarity: “YOUR TEXT”

Construct a table for the following passage. In the left column should appear the topic information in each sentence. In the right column should appear the information in the stress position of each sentence. Then, analyze the table for shortcomings and propose a new table of topic and stress positions that has better logical flow. Finally, rewrite the passage using your new table. Here is the text: “YOUR TEXT”

Advanced

More on topic and stress positions: G open and Swan 1990

95 of 105

Quadrant 2: Clear thinking, not enough text

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

-templates + text generation

-make time, space, goals

96 of 105

AI strategies: Quadrant 2

Choose a template that suits your text

Use AI to help generate a template

Your task is to assemble the following sentences into a {paragraph/section/etc}. Your response should retain as much of the original language as possible but improve the clarity. You should use appropriate scientific terminology where relevant, while also ensuring that any technical terms are explained clearly. Your response should retain all necessary details from the original sentence, such as numbers or specific names. You should maintain a formal tone throughout. Here is the text: “YOUR TEXT”

Advanced

97 of 105

Non-AI strategy: Quadrant 2

Make time - Pre-commit to writing times each week

Photo by Phil

Make goals - Set an attainable goal in advance of each session and check it off when you finish

Make space - Put your phone away, close email and social media, turn notifications off; tell family and coworkers not to disturb you

Photo by Agê Barros

Photo by Muzammil Soorma

98 of 105

Quadrant 3: Muddy thinking, not enough text

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

-idea generation

-free writing

-focused gathering

99 of 105

AI strategies: Quadrant 3

Generate ideas for a discussion section for this paper: “YOUR TEXT”

I answered a series of questions that served as prompts for me to get started writing a discussion section in a scientific manuscript. I also have a discussion section structure that I'd like to use. I'm going to give you that structure and then give you my answers, and I want you to outline a discussion section paragraph by paragraph. Retain detail and include analysis. Suggest places where citations should appear to support the argument. “YOUR TEXT”

Structure

voice-to-text

ChatGPT

template

Advanced

100 of 105

Non-AI strategy: Focused gathering

Gather and read materials, but with purpose

How and with what evidence do I want to prove my points?

What gaps do I have in my knowledge?

@artifexx

101 of 105

Quadrant 4: Muddy thinking, too much text

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

-templates + text revisions

-reverse outlining

102 of 105

AI strategies: Quadrant 4

I have a draft of a discussion section from a scientific manuscript. I also have a discussion section structure that I'd like to use. I'm going to give you my draft and that structure. I want you to provide an outline that reorganizes my text according to the new structure. Here is the draft: “YOUR TEXT”. Here is the structure: “YOUR TEXT”

Reverse outline the following passage of text. Create a table with the left column explaining the main point of each paragraph, and a right column explaining how the paragraph advances the argument. Analyze the structure of the passage, and propose any structural revisions. Here is the text: “YOUR TEXT”

Advanced

More on reverse outlining: Reverse outline - AI writing Columbia [20230502]

103 of 105

Spaces in the writing process

too much text

not enough text

clear thinking

muddy thinking

Adapted from Liz Neeley / Liminal Creations

-templates + text revisions

-reverse outlining

-idea generation

-free writing

-focused gathering

-grammar

-clarity

-logical flow

-templates + text generation

-make time, space, goals

1 of 105

2 of 105

3 of 105

4 of 105

5 of 105

6 of 105

7 of 105

8 of 105

9 of 105

10 of 105

11 of 105

12 of 105

13 of 105

14 of 105

15 of 105

16 of 105

17 of 105

18 of 105

19 of 105

20 of 105

21 of 105

22 of 105

23 of 105

24 of 105

25 of 105

26 of 105

27 of 105

28 of 105

29 of 105

30 of 105

31 of 105

32 of 105

33 of 105

34 of 105

35 of 105

36 of 105

37 of 105

38 of 105

39 of 105

40 of 105

41 of 105

42 of 105

43 of 105

44 of 105

45 of 105

46 of 105

47 of 105

48 of 105

49 of 105

50 of 105

51 of 105

52 of 105

53 of 105

54 of 105

55 of 105

56 of 105

57 of 105

58 of 105

59 of 105

60 of 105

61 of 105

62 of 105

63 of 105

64 of 105

65 of 105

66 of 105

67 of 105

68 of 105

69 of 105

70 of 105

71 of 105

72 of 105

73 of 105

74 of 105

75 of 105

76 of 105

77 of 105

78 of 105

79 of 105

80 of 105