Throughout 2023, it felt like the drama never let up. From Elon Musk's nonstop shenanigans to the constant launches in the generative AI race, the last twelve months was packed with news. Thankfully, it wasn't all bad, and this year saw more winners than before. There were clear frontrunners, like Threads and AI, but we also saw surprises like Apple's Vision Pro headset and the iPhone maker finally embracing several open standards. Of all the things that happened this year, here's the Engadget team's list of tech's biggest winners in 2023.

If you're waiting for the hubbub over generative AI to die down, maybe pull up a chair. The buzz around artificial intelligence shows no signs of quieting--a fact that will become all too obvious at this year's CES. CES, the consumer electronics industry's largest annual gathering in the US, is returning to Las Vegas on January 9. It is a massive, four-day-long bustling bazaar of tech, with expo halls filled to the brim with new gadgets, hopeful startups, and prototypes that reach for the stars. CES is a trade show where sales and distribution deals are inked, where concept cars roll through crowded streets, and where tech journalists and showgoers alike wander the floors looking for the standout new products.

The artificial intelligence (AI) industry began 2023 with a bang as schools and universities struggled with students using OpenAI's ChatGPT to help them with homework and essay writing. Less than a week into the year, New York City Public Schools banned ChatGPT – released weeks earlier to enormous fanfare – a move that would set the stage for much of the discussion around generative AI in 2023. As the buzz grew around Microsoft-backed ChatGPT and rivals like Google's Bard AI, Baidu's Ernie Chatbot and Meta's LLaMA, so did questions about how to handle a powerful new technology that had become accessible to the public overnight. In March, a group of more than 1,000 signatories, including Apple co-founder Steve Wozniak and billionaire tech entrepreneur Elon Musk, called for a pause in the development of more advanced AI in light of its "profound risks to society and humanity". While a pause did not happen, governments and regulatory authorities began rolling out new laws and regulations to set guardrails on the development and use of AI.

Modern generative models exhibit unprecedented capabilities to generate extremely realistic data. However, given the inherent compositionality of the real world, reliable use of these models in practical applications requires that they exhibit the capability to compose a novel set of concepts to generate outputs not seen in the training data set. Prior work demonstrates that recent diffusion models do exhibit intriguing compositional generalization abilities, but also fail unpredictably. Motivated by this, we perform a controlled study for understanding compositional generalization in conditional diffusion models in a synthetic setting, varying different attributes of the training data and measuring the model's ability to generate samples out-of-distribution. Our results show: (i) the order in which the ability to generate samples from a concept and compose them emerges is governed by the structure of the underlying data-generating process; (ii) performance on compositional tasks exhibits a sudden "emergence" due to multiplicative reliance on the performance of constituent tasks, partially explaining emergent phenomena seen in generative models; and (iii) composing concepts with lower frequency in the training data to generate out-of-distribution samples requires considerably more optimization steps compared to generating in-distribution samples. Overall, our study lays a foundation for understanding emergent capabilities and compositionality in generative models from a data-centric perspective.

Diffusion Models (DMs) are state-of-the-art generative models that learn a reversible corruption process from iterative noise addition and denoising. They are the backbone of many generative AI applications, such as text-to-image conditional generation. However, recent studies have shown that basic unconditional DMs (e.g., DDPM [16] and DDIM [52]) are vulnerable to backdoor injection, a type of output manipulation attack triggered by a maliciously embedded pattern at model input. This paper presents a unified backdoor attack framework (VillanDiffusion) to expand the current scope of backdoor analysis for DMs. Our framework covers mainstream unconditional and conditional DMs (denoising-based and score-based) and various training-free samplers for holistic evaluations. Experiments show that our unified framework facilitates the backdoor analysis of different DM configurations and provides new insights into caption-based backdoor attacks on DMs.

The prompt asked for a 1,200 word article (a number it undercut by quite a margin) that explored both the potential negative and positive outcomes of the technology. We then asked it to include real world examples, which is apparently beyond its capabilities. We also asked it to include a section on the recent Sam Altman debacle which, as you will soon read, was also not a topic it was particularly capable at describing. Below is the unedited output with light changes for formatting. Generative Artificial Intelligence (AI) has emerged as a powerful force, reshaping the technological landscape with its ability to create content autonomously.

Let's take a look at how OpenAI and its chatbot have impacted consumer electronics in 2023 and where they might lead the industry in the new year. "Meteoric" doesn't do justice to OpenAI's rise this year. The company released ChatGPT on November 30, 2022. Within five days, the program had passed 1 million users; by January, 100 million people a month were logging on to use it. It took Facebook four and a half years to reach those sorts of engagement numbers.

If 2022 was the year the generative AI boom started, 2023 was the year of the generative AI panic. Just over 12 months since OpenAI released ChatGPT and set a record for the fastest-growing consumer product, it appears to have also helped set a record for fastest government intervention in a new technology. The US Federal Elections Commission is looking into deceptive campaign ads, Congress is calling for oversight into how AI companies develop and label training data for their algorithms, and the European Union passed its new AI Act with last-minute tweaks to respond to generative AI. But for all the novelty and speed, generative AI's problems are also painfully familiar. OpenAI and its rivals racing to launch new AI models are facing problems that have dogged social platforms, that earlier era-shaping new technology, for nearly two decades.

The New York Times has sued Microsoft and OpenAI for using its content to help develop artificial intelligence services, in a sign of the increasingly fraught relationship between the media and a technology that could upend the news industry. The Times didn't specify its monetary demands. OpenAI has faced criticism for scraping text widely from the web to train its popular chatbot since it debuted a year ago. While it has been sued by prominent authors, this is the first challenge to its practices by a major media organization. The startup has sought licensing deals with publishers, much like Alphabet's Google and Meta Platforms' Facebook have done in recent years.

The New York Times has sued OpenAI and Microsoft over the use of its content to train generative artificial intelligence and large-language model systems, a move that could see the company receive billions of dollars in damages. The lawsuit contains an appeal to the "vital" importance of the Times's independent journalism to democracy, arguing that it is "increasingly rare and valuable". The publisher's lawsuit is the latest in a string of similar cases, including one brought by more than a dozen authors in September targeting the company for its use of their writing. Language learning models have faced increasing scrutiny since they exploded in popularity in the past year, with news outlets in particular concerned that the tools will spread misinformation attributed to them and utilize their content with no incentive to click through to the original source. ChatGPT launched in November 2022 and amassed 100 million users in just two months.