William S. Saunders

We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics.

Most tumor cells are characterized by increased genomic instability and chromosome segregational defects, often associated with hyperamplification of the centrosome and the formation of multipolar spindles. However, extra centrosomes do not always lead to multipolarity. Here, we describe a process of centrosomal clustering that prevented the formation of multipolar spindles in noncancer cells. Noncancer cells needed to overcome this clustering mechanism to allow multipolar spindles to form at a high frequency. The microtubule motor cytoplasmic dynein was a critical part of this coalescing machinery, and in some tumor cells overexpression of the spindle protein NuMA interfered with dynein localization, promoting multipolarity.

Oral squamous cell carcinomas are characterized by complex, often near-triploid karyotypes with structural and numerical variations superimposed on the initial clonal chromosomal alterations. We used immunohistochemistry combined with classical cytogenetic analysis and spectral karyotyping to investigate the chromosomal segregation defects in cultured oral squamous cell carcinoma cells. During division, these cells frequently exhibit lagging chromosomes at both metaphase and anaphase, suggesting defects in the mitotic apparatus or kinetochore. Dicentric anaphase chromatin bridges and structurally altered chromosomes with consistent long arms and variable short arms, as well as the presence of gene amplification, suggested the occurrence of breakage-fusion-bridge cycles. Some anaphase bridges were observed to persist into telophase, resulting in chromosomal exclusion from the reforming nucleus and micronucleus formation. Multipolar spindles were found to various degrees in the oral squamous cell carcinoma lines. In the multipolar spindles, the poles demonstrated different levels of chromosomal capture and alignment, indicating functional differences between the poles. Some spindle poles showed premature splitting of centrosomal material, a precursor to full separation of the microtubule organizing centers. These results indicate that some of the chromosomal instability observed within these cancer cells might be the result of cytoskeletal defects and breakage-fusion-bridge cycles.

We have used a single-gene deletion mutant bank to identify the genes required for meiosis and sporulation among 4323 nonessential Saccharomyces cerevisiae annotated open reading frames (ORFs). Three hundred thirty-four sporulation-essential genes were identified, including 78 novel ORFs and 115 known genes without previously described sporulation defects in the comprehensive Saccharomyces Genome (SGD) or Yeast Proteome (YPD) phenotype databases. We have further divided the uncharacterized sporulation-essential genes into early, middle, and late stages of meiosis according to their requirement for IME1 induction and nuclear division. We believe this represents a nearly complete identification of the genes uniquely required for this complex cellular pathway. The set of genes identified in this phenotypic screen shows only limited overlap with those identified by expression-based studies.