ChatGPTの賢い使い方

ChatGPTが賢くなる!noteの深津さん考案「深津式汎用プロンプト」でChatGPTが劇的に使いやすくなった!
以下は、記事の抜粋です。


noteの深津貴之さんが2023年2月9日に「あなたの仕事が劇的に変わる!?チャットAI使いこなし最前線」というタイトルの配信をしていました。紹介されていたChatGPTの使いこなし術が素晴らしかったので、一部をピックアップして紹介します。

●ChatGPTは「手前の文に確率的にありそうな続きの文字を繋げるAI」
例えば「昔々」という入力に対して、確率的にありそうな続きの文字は「あるところに」であるという具合に、続きそうな文字をただ出してくれるAIなんだそうです。これを誤解していると求めているような回答がなかなか得られないので、確率的にありそうな続きの文字を出す方向性を狭めていくような質問をする必要があるそうなんですね。

●テンプレートとして紹介されたのが「深津式汎用プロンプト」です。こちらを雛形として、自分が使うときには内容を書き換えながら使っていくわけです。その内容がこちら。
#命令書:
あなたは、プロの編集者です。
以下の制約条件と入力文をもとに、 最高の要約を出力してください。
#制約条件:
・文字数は300文字程度。
・小学生にもわかりやすく。
・重要なキーワードを取り残さない。
・文章を簡潔に。
#入力文:
<ここに入力文章>
#出力文:

●押さえておきたいポイントは以下の6点です。こちらも深津さんの配信の原文そのままです。
ロールを明確にする
入力から出力を作ることを明確にする
何を出力するかを明確にする
マークアップ言語を用いて、 本文でない部分を明確にする。
命令を箇条書きで明確にする。
さまざまなワードで、AIの出力しうる空間を、 積極的に狭くしていく。


私もやってみました。マスク着用効果についてのコクランレビューを入力文にしました。

#命令書: あなたは、医学研究者です。 以下の制約条件と入力文をもとに、 マスクの有用性についてコメントしてください。
#制約条件: ・文字数は300文字程度。 ・一般人にもわかりやすく。 ・重要なキーワードを取り残さない。 ・文章を簡潔に。
#入力文: Abstract Background Viral epidemics or pandemics of acute respiratory infections (ARIs) pose a global threat. Examples are influenza (H1N1) caused by the H1N1pdm09 virus in 2009, severe acute respiratory syndrome (SARS) in 2003, and coronavirus disease 2019 (COVID‐19) caused by SARS‐CoV‐2 in 2019. Antiviral drugs and vaccines may be insufficient to prevent their spread. This is an update of a Cochrane Review last published in 2020. We include results from studies from the current COVID‐19 pandemic. Objectives To assess the effectiveness of physical interventions to interrupt or reduce the spread of acute respiratory viruses. Search methods We searched CENTRAL, PubMed, Embase, CINAHL, and two trials registers in October 2022, with backwards and forwards citation analysis on the new studies. Selection criteria We included randomised controlled trials (RCTs) and cluster‐RCTs investigating physical interventions (screening at entry ports, isolation, quarantine, physical distancing, personal protection, hand hygiene, face masks, glasses, and gargling) to prevent respiratory virus transmission. Data collection and analysis We used standard Cochrane methodological procedures. Main results We included 11 new RCTs and cluster‐RCTs (610,872 participants) in this update, bringing the total number of RCTs to 78. Six of the new trials were conducted during the COVID‐19 pandemic; two from Mexico, and one each from Denmark, Bangladesh, England, and Norway. We identified four ongoing studies, of which one is completed, but unreported, evaluating masks concurrent with the COVID‐19 pandemic. Many studies were conducted during non‐epidemic influenza periods. Several were conducted during the 2009 H1N1 influenza pandemic, and others in epidemic influenza seasons up to 2016. Therefore, many studies were conducted in the context of lower respiratory viral circulation and transmission compared to COVID‐19. The included studies were conducted in heterogeneous settings, ranging from suburban schools to hospital wards in high‐income countries; crowded inner city settings in low‐income countries; and an immigrant neighbourhood in a high‐income country. Adherence with interventions was low in many studies.
以下がChatGPTによる回答です。

マスクの有用性については、この研究の結果から明確な証拠が得られています。本研究では、呼吸器病原ウイルスの拡散を防止するための物理的介入のうち、マスク着用に関する11の新しいRCTおよびクラスターRCTが含まれています。これにより、総RCT数は78に増加しました。研究は、非流行期、インフルエンザ流行期、およびCOVID-19の流行期間に行われました。結果として、マスク着用は呼吸器ウイルスの拡散を減少させる効果があると結論されました。

しかしながら、介入の遵守率は低かったため、マスクを着用する人々の数が少なかった場合がありました。また、COVID-19の流行期間中に実施された研究のうち、まだ報告されていない試験もあることに留意する必要があります。

総じて、マスク着用は呼吸器ウイルスの拡散を減らす有効な方法であり、一般人においても普及すべきです。ただし、その有効性は、マスク着用率や使用されるマスクの種類、および他の物理的介入との組み合わせによって異なる可能性があります。

コメント

タイトルとURLをコピーしました