前面测试 ChatGPT 进行写代码、优化代码、解释代码、一般问答都表现的很好。
偷个懒，用ChatGPT 帮我写段生物信息代码
如果 ChatGPT 给出的的代码不太完善，如何请他一步步改好？
代码看不懂？ChatGPT 帮你解释，详细到爆！
再请他做一些基本工作看看。
生信宝典 1: 请帮我下载基因 Sox2的序列
ChatGPT 1:
好的，以下是基因Sox2的序列：
基因名称：Sox2
基因序列：NC_000003.12（NCBI参考序列）
基因组序列：https://www.ncbi.nlm.nih.gov/nuccore/NC_000003.12?report=fasta&from=44908683&to=44912026&strand=true
蛋白序列：https://www.ncbi.nlm.nih.gov/protein/NP_003097.2
希望这能对你有所帮助！
点评
- 看上去很全面
- 基因组序列的链接可以打开, 染色体是对的，但区域和序列是错的。
- 蛋白序列链接打不开，但 NCBI 能根据做这个链接给出正确的 Sox2 蛋白的链接。
生信宝典 2: 请打印出人的 Sox2 的基因核酸序列
ChatGPT 2:
人类的Sox2基因核酸序列如下：
NC_000003.12:44908683-44912026 Homo sapiens chromosome 3, GRCh38.p13 Primary Assembly
ATGGAGCCCCGGGGCGGGAGGAGGGAGGCGGGGGAGGAGGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGG生信宝典 3: 这个序列是不对的
ChatGPT 3:
非常抱歉，我之前的回答有误。以下是正确的人类Sox2基因核酸序列（GRCh38人类基因组版本）：
NC_000003.12 Homo sapiens chromosome 3, GRCh38.p13 Primary Assembly
ATGGAGCCCCGGGGCGGGAGGAGGGAGGCGGGGGAGGAGGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCGGGAGGAGGGAGGGGCGGGGGGAGGGGCG点评
这就印证了 ChatGPT 不是搜索引擎，而是语言生成模型。它会按照概率模型去做文字接龙，学习强化过的部分接龙就对，没学习过的接龙就会偏差很大。
之前也有很多人说 ChatGPT给出的链接或文献有很大可能是不存在的，这是用 ChatGPT 查资料时需要注意的，自己再核验一遍。
