Categories
News

Epoch AI lanza el punto de referencia FrontierMath AI para probar las capacidades de los modelos de IA

[ad_1]

Epoch AI, un instituto de investigación con sede en California, ha lanzado un nuevo programa inteligencia artificial (AI) récord la semana pasada. El nuevo estándar de IA, llamado FrontierMath, prueba la capacidad de los modelos de lenguaje grandes (LLM) para refactorizar y resolver problemas matemáticos. La empresa de IA afirma que los estándares matemáticos actuales no son muy útiles debido a factores como la contaminación de datos y los modelos de IA que obtienen puntuaciones demasiado altas en ellos. Epoch AI afirma que incluso los principales titulares de LLM obtuvieron menos del dos por ciento en el nuevo punto de referencia.

Epoch AI lanza el estándar FrontierMath

en un correo En X (anteriormente conocido como Twitter), la empresa de inteligencia artificial explicó que ha colaborado con más de 60 matemáticos para crear cientos de activos y problemas matemáticos inéditos. Epoch AI afirma que incluso los matemáticos podrían tardar horas en resolver estas preguntas. La razón detrás del desarrollo del nuevo estándar se citó como las limitaciones de los estándares existentes como GSM8K y MATH, donde los modelos de IA generalmente obtienen una puntuación alta.

La empresa afirmó que las altas puntuaciones obtenidas por los LLM se debían en gran medida a la contaminación de datos. Esto significa que las preguntas ya se han introducido de una forma u otra en los modelos de IA, lo que permite que las preguntas se resuelvan fácilmente.

FrontierMath resuelve el problema al incluir nuevos problemas que son únicos y no están publicados en ninguna parte, mitigando los riesgos asociados con la contaminación de datos. Además, el estándar incluye una amplia gama de preguntas que incluyen problemas computacionales intensivos en teoría de números, análisis real y geometría algebraica, así como temas como la teoría de grupos de Zermelo-Fränkel. La compañía de inteligencia artificial dice que todas las preguntas son “a prueba de conjeturas”, lo que significa que no se pueden resolver accidentalmente sin pensar detenidamente.

Epoch AI destacó que para medir la eficiencia de la IA, se deben crear criterios para la resolución creativa de problemas, ya que la IA debe sustentar el pensamiento en múltiples pasos. En particular, muchos expertos de la industria creen que los estándares actuales no son suficientes para medir adecuadamente el progreso de un modelo de IA.

Respuesta a la nueva norma en A correoNoam Brown, el investigador de OpenAI que estuvo detrás del modelo o1 de la compañía, dio la bienvenida al nuevo estándar y dijo: “Me encantaría ver una nueva evaluación con bajas tasas de éxito para modelos paramétricos”.

Para lo último Noticias de tecnología y ReseñasSiga Gadgets 360 en incógnita, Facebook, WhatsApp, Temas y noticias de google. Para ver los últimos vídeos sobre gadgets y tecnología, suscríbete a nuestro canal. canal de youtube. Si quieres saber todo sobre los top influencers, sigue nuestra web ¿Quién es ese 360? en Instagram y YouTube.


Poco X7 Pro podría ser el primer teléfono inteligente que se envíe con HyperOS 2 de Xiaomi en India



Las opciones de color del iQOO 13 se revelaron antes de su lanzamiento en India el 3 de diciembre.



[ad_2]

Source Article Link

Categories
Life Style

the Anthropocene is not an epoch, despite protest over vote

[ad_1]

Plastic bottles washed up on shore of a Carribean islanda`

Some geoscientists argue that humans have transformed the planet with plastic trash, radioactive debris and fossil-fuel emissions, among other things — and that the changes should be recognized with a new geological epoch.Credit: Mark Meredith/Getty

A high-profile battle over whether to designate the ‘Anthropocene’ as a new geological epoch has come to an end. On 20 March, the International Union of Geological Sciences (IUGS) — the final arbiter in the matter — announced it is upholding a decision made earlier this month by a group of geoscientists. That group voted on 4 March to reject a proposal that would have established the current era, in which humans are altering the planet, as a formal epoch in Earth’s geological timetable.

The IUGS decision effectively terminates a dramatic challenge to that earlier vote: the chair and a vice-chair of the Subcommission on Quaternary Stratigraphy (SQS), which held the vote, said it was illegitimate. Among other things, they said that 11 of the 16 SQS members who voted on the Anthropocene proposal were ineligible because they have been members of the subcommission for too long.

In a statement, the IUGS called the 4 March vote and its appeal “a difficult process” that was conducted “fully in accordance with the statutory requirements”. Stanley Finney, a stratigrapher at California State University, Long Beach, who is the IUGS secretary-general, told Nature that it was longstanding practice among these subcommissions to allow members who had overextended their terms to vote anyway. “You can’t just throw them off if you want something done,” he says.

The IUGS is the parent organization for the International Commission on Stratigraphy (ICS), to which the SQS belongs. “There is no further supreme court one can go to,” says Jan Zalasiewicz, a palaeontologist at the University of Leicester, UK, who is the SQS chair who protested the subcommission’s vote. “I have no immediate plans for a challenge.”

A tangled quest

The controversy underscores the long-running quest to bring the Anthropocene proposal to a vote, and the tangle of international geological organizations involved. In 2009, the SQS set up an Anthropocene working group to assess whether the current era of human-induced change should be codified as a new ‘stratigraphic unit’ in the geological time scale. After 15 years of discussion and exploration, the working group submitted its proposal last October, arguing that a new epoch should be established. Its start, the group said, should be marked by plutonium residue from hydrogen-bomb tests in 1952 appearing in Earth’s geology.

Under ICS rules, such a proposal would normally be discussed for a 30-day period and then voted on for another 30 days. Zalasiewicz says he and vice-chair Martin Head, a stratigrapher at Brock University in St Catharines, Canada, recused themselves from moderating the discussion because of conflicts of interest from their earlier participation in the Anthropocene working group. When the discussion period ended and other SQS members moved to vote on the proposal, Zalasiewicz and Head objected, saying that it had not been given serious consideration and that the vote was rushed.

Voting began in early February and ended on 4 March, with four SQS members voting in favour of establishing an Anthropocene epoch and 12 voting against it. Three people abstained and three did not vote, including Zalasiewicz and Head. The results of the vote were then approved by the full ICS and, as of today, the IUGS.

Participation on ICS subcommissions, which deal with different geological periods from the ancient to the modern, typically happens in four-year terms. Anyone who has been a member of a subcommission for more than 12 years is no longer a voting member, according to ICS statutes. Zalasiewicz says that this restriction applies to him, as well as to many other SQS members — and says that for this reason, the 4 March vote is illegitimate.

Stricter compliance

The discussion has prompted other ICS subcommissions to re-examine their membership rosters for people who might have passed the 12-year limit and thus are no longer eligible to vote, says Philip Gibbard, a geologist at the University of Cambridge, UK. More broadly, the IUGS has been working towards refreshing its committee membership more frequently, to increase gender, racial and geographic equity, Finney says.

David Harper, a palaeontologist at Durham University, UK, and current chair of the ICS, says that the commission will be enforcing stricter compliance on term limits going forward.

For his part, Zalasiewicz says he has been asked to step down as SQS chair and does not expect to be part of any ICS group going forward. He and other Anthropocene-epoch advocates are likely to continue their campaign in other venues, he says: “Another means will have to be found” to codify the Anthropocene as a concept outside of the official geological timescale.

Regardless of there being no formal Anthropocene epoch, the term will continue to be used in broad popular and scientific usage as the era of human-induced change. “As such, it will remain an invaluable descriptor in human-environment interactions,” the IUGS says.

[ad_2]

Source Article Link

Categories
Life Style

Geologists reject the Anthropocene as Earth’s new epoch — after 15 years of debate

[ad_1]

After 15 years of discussion and exploration, a committee of researchers has decided that the Anthropocene — generally understood to be the age of irreversible human impacts on the planet — will not become an official epoch in Earth’s geological timeline. The ruling, first reported by The New York Times, is meant to be final, but is being challenged by two leading members of the committee that ran the vote.

Twelve members of the international Subcommission on Quaternary Stratigraphy (SQS) voted against the proposal to create an Anthropocene epoch, and only four voted for it. That would normally constitute an unqualified defeat, but a dramatic challenge has arisen from the chair of the SQS, palaeontologist Jan Zalasiewicz at the University of Leicester, UK, and one of the group’s vice-chairs, stratigrapher Martin Head at Brock University in St Catharines, Canada.

In a 6 March press statement, they said that they are asking for the vote to be annulled. They added that “the alleged voting has been performed in contravention of the statutes of the International Commission on Stratigraphy” (ICS), including statutes governing the eligibility to vote. Zalasiewicz told Nature that he couldn’t comment further just yet, but that neither he nor Head had “instigated the vote or agreed to it, so we are not responsible for procedural irregularities”.

The SQS is a subcommittee of the ICS. Normally, there would be no appeals process for a losing vote. ICS chair David Harper, a palaeontologist at Durham University, UK, had confirmed to Nature before the 6 March press statement that the proposal “cannot be progressed further”. Proponents could put forward a similar idea in the future.

If successful, the proposal would have codified the end of the current Holocene epoch, which has been going on since the end of the last ice age 11,700 years ago, and set the start of the Anthropocene in the year 1952. This is when plutonium from hydrogen-bomb tests showed up in the sediment of Crawford Lake near Toronto, Canada, a site chosen by some geologists to be designated as a ‘golden spike’ as capturing a pristine record of humans’ impact on Earth. Other signs of human influence in the geological record include microplastics, pesticides and ash from fossil-fuel combustion.

But pending the resolution of the challenge, the lake and its plutonium residue won’t get a golden spike. Selecting one site as such a marker “always felt a bit doomed, because human impacts on the planet are global”, says Zoe Todd, an anthropologist at Simon Fraser University in Burnaby, Canada. “This is actually an invitation for us to completely rethink how we define what the world is experiencing.”

A cultural concept

Although the Anthropocene probably will not be added to the geological timescale, it remains a broad cultural concept already used by many to describe the era of accelerating human impacts, such as climate change and biodiversity loss. “We are now on a fundamentally unpredictable planet in ways that we have not experienced for the last 12,000 years,” says Julia Adeney Thomas, a historian at the University of Notre Dame, in Indiana. “That understanding of the Anthropocene is crystal clear.”

The decision to reject the designation was made public through The New York Times on 5 March, after the SQS had concluded its month-long voting process, but before committee leaders had finalized discussions and made an official announcement. Philip Gibbard, a geologist at the University of Cambridge, UK, who is on the SQS, says that the crux of the annulment challenge is that Zalasiewicz and Head objected to the voting process kicking off on 1 February. The rest of the committee wanted to move forward with a vote and did so according to SQS rules, Gibbard says. “There’s a lot of sour grapes going on here,” he adds.

Had the proposal made it through the SQS, it would have needed to clear two more hurdles: first, a ratification vote by the full stratigraphic commission, and then a final one in August at a forum of the International Union of Geological Sciences.

Frustrated by defeat

Some of those who helped to draw up the proposal, through an Anthropocene working group commissioned by the SQS, are frustrated by the apparent defeat. They had spent years studying a number of sites around the world that could represent the start of a human-influenced epoch. They performed fresh environmental analyses on many of the sites, including studying nuclear debris, fossil-fuel ash and other markers of humans’ impact in geological layers, before settling on Crawford Lake.

“We have made it very clear that the planet we’re living on is different than it used to be, and that the big tipping point was in the mid-twentieth century,” says Francine McCarthy, a micropalaeontologist at Brock University who led the Crawford Lake proposal1. Even though the SQS has rejected it, she says she will keep working to highlight the lake’s exceptionally preserved record of human activities. “Crawford Lake is just as great a place as it ever was.”

“To be honest, I am very disappointed with the SQS outcome,” says working-group member Yongming Han, a geochemist at the Institute of Earth Environment of the Chinese Academy of Sciences in Xi’an. “We all know that the planet has entered a period in which humans act as a key force and have left indisputable stratigraphic evidences.”

For now, the SQS and the ICS will sort out how to handle Zalasiewicz and Head’s request for a vote annulment. Meanwhile, scientific and public discussions about how best to describe the Anthropocene continue.

One emerging argument is that the Anthropocene should be defined as an event in geological history — similar to the rise of atmospheric oxygen just over two billion years ago, known as the Great Oxidation Event — but not as a formal epoch2. This would make more sense because geological events unfold as transformations over time, such as humans industrializing and polluting the planet, rather than as an abrupt shift from one state to another, says Erle Ellis, an ecologist at the University of Maryland Baltimore County in Baltimore. “We need to think about this as a broader process, not as a distinct break in time,” says Ellis, who resigned from the Anthropocene working group last year because he felt it was looking at the question too narrowly.

This line of thinking played a part in at least some of the votes to reject the idea of an Anthropocene epoch. Two SQS members told Nature they had voted down the proposal in part because of the long and evolving history of human impacts on Earth.

“By voting ‘no’, they [the SQS] actually have made a stronger statement,” Ellis says: “that it’s more useful to consider a broader view — a deeper view of the Anthropocene.”

[ad_2]

Source Article Link